El Paso Data Center Readiness Audit — verdict

Run date: 2026-09-03
Overall verdict: THEORY SUPPORTED
Rules checksum: fb45ba1c172b0239… (config/thresholds.yaml)
Findings checksum: d3ecec0f2e6e5d7a…

Thresholds were fixed in PREREGISTRATION.md before collection. If the rules checksum above does not match that document's rules, this report was produced under different rules and is invalid.

P1 confirmed, and 5 of 5 remaining propositions confirmed (threshold 4).

Theory under test

El Paso is severely under-delivering in having workforce and contracting ready for data center buildouts.

Measured against the Meta / BlackRock El Paso data center campus: 1000 MW, operational 2028, 4,000 construction workers at peak, 300 permanent jobs.

Summary

🔴 supports the theory · 🟢 counts against it · ⚪ the record cannot say

Findings

P1 — Demand is real, scheduled, and sized

Verdict: CONFIRMED

A 1000 MW campus is on the public record with a dated load ramp in a PUCT filing and a stated peak construction workforce of 4,000. The demand signal is not speculative.

Pre-registered rule: CONFIRMED if capacity and dates appear in at least one regulatory filing or binding corporate announcement.

Evidence

• Meta / BlackRock El Paso data center campus: 1000 MW final capacity, operational 2028. _{crosswalk.yaml:demand}
• Peak construction workforce 4,000; permanent jobs 300. _{crosswalk.yaml:demand}
• Load ramp 2026: 220 MW — original EPE request, PUCT-approved rates _{2026 · PUCT filing}
• Load ramp 2027: 440 MW — projected demand per EPE PUCT filing _{2027 · PUCT filing}
• Load ramp 2029: 1000 MW — projected demand per EPE PUCT filing _{2029 · PUCT filing}
• Bridge generation McCloud Generation: 366 MW across 813 units. Modular gas generators adjacent to site; exclusive to the data center for an initial five-year bridge period from 2027; separate from EPE transmission system. _{PUCT filing}
• Source: Meta newsroom, 'Meta Announces New Strategic Venture with BlackRock to Develop Data Center in El Paso', 2026-07-21 _primary
• Source: El Paso Electric filing, Public Utility Commission of Texas (McCloud Generation / Wurldwide LLC load) _primary
• Source: Bloomberg reporting on the $12B bond sale, 2026-07-21 _primary

Caveats

• P1 is a precondition, not a finding against El Paso. It establishes the denominator the remaining propositions are measured against.
• Peak construction headcount is the figure released by the developer. If it is revised down materially, P2 must be recomputed.

P2 — Local skilled-trades stock cannot serve peak demand

Verdict: CONFIRMED

One project's peak electrician demand equals 53%-71% of every job in El Paso County, TX's electrical contracting sector. The pre-registered threshold is 25%.

Pre-registered rule: CONFIRMED if peak trade demand exceeds 25% of county trade employment at the LOW end of the trade-share range (30%); REFUTED if below that share at the HIGH end (40%).

Evidence

• El Paso County, TX employment in NAICS 23821 (Electrical and wiring contractors), 2025Q4: 2,263 jobs. _{2025Q4 · NAICS 23821}
• Across 173 establishments — average 13.1 employees per establishment. _{2025Q4 · NAICS 23821}
• Peak electrician demand on this project: 1,200-1,600 workers (30%-40% of a 4,000-worker peak). _{industry trade-share range}
• Ratio of peak demand to existing county stock: 53%-71%. _derived

Caveats

• QCEW counts jobs covered by unemployment insurance in firms based in the county. It excludes the self-employed and counts travelling crews at their employer's location, so it measures the local employer base rather than every worker who might appear on a site.
• The denominator is the whole sector, including workers already committed to residential and commercial jobs. It is a ceiling on available local labour, not a pool of idle workers.
• Trade share is a published industry range, not an El Paso-specific measurement. Both ends are reported for that reason.
• TWO PEAK FIGURES EXIST AND THIS MATTERS. The City's 2023-12-04 council presentation states 'Expect 1,000+ skilled trade workers at peak construction' for the INITIAL PHASE (minimum $800M, ~800k sq ft, one of up to five phases). Meta's 2026-07-21 announcement states 4,000+ at peak for the full ~1 GW campus. This audit uses 4,000 because it is the current figure for the full campus, but the distinction is a phase distinction, not a contradiction. On the 2023 initial-phase figure the ratio would be 13-18%, which is BELOW the 25% threshold and would REFUTE P2 for that phase alone. Anyone citing this finding must say which scope they mean.

P3 — The training pipeline is not scaled to the ramp

Verdict: CONFIRMED

El Paso County, TX produces 32 electrician completers a year against an annual incremental need of 400-533 — 6%-8% of demand, against a 20% threshold.

Pre-registered rule: P3a CONFIRMED if annual local completers are below 20% of annual incremental demand. P3b CONFIRMED if no step-change of 15%+ appears in any measured series after 2026-07-21.

Evidence

• El Paso County, TX, 2023-24: 32 completers in electrician programs (CIP 46.0302). _{2023-24 · IPEDS C_A}
• Across all tracked trade programs (electrician, HVAC, plumbing, industrial electronics, industrial maintenance, EE tech, network admin): 130 completers. _{2023-24 · IPEDS C_A}
• Annual incremental electrician demand: 400-533 per year (30%-40% of a 4,000-worker peak, spread over 3 years). _derived
• Local electrician output covers 6%-8% of annual incremental demand. _derived
• Timing test — electrical trade employment: 2,309 (2023Q4) -> 2,263 (2025Q4), -2.0% over the 8 quarters since the 2023-12-04 council decision. _{2023Q4->2025Q4 · QCEW}
• Timing test — electrical trade employment establishments: 181 (2023Q4) -> 173 (2025Q4), -4.4% over the 8 quarters since the 2023-12-04 council decision. _{2023Q4->2025Q4 · QCEW}
• Timing test — building equipment trade employment: 4,541 (2023Q4) -> 4,678 (2025Q4), +3.0% over the 8 quarters since the 2023-12-04 council decision. _{2023Q4->2025Q4 · QCEW}
• Timing test — building equipment trade employment establishments: 433 (2023Q4) -> 418 (2025Q4), -3.5% over the 8 quarters since the 2023-12-04 council decision. _{2023Q4->2025Q4 · QCEW}
• IPEDS cannot yet show a response: its latest period is 2023-24. The QCEW series above can, and cover 2023-12-04 to the present. _{reporting lag}

Caveats

• IPEDS covers Title IV institutions only. Records produced by Workforce Solutions Borderplex on 2026-08-20 confirm at least ten registered apprenticeship programmes or sponsors in El Paso trades — the IBEW 583/NECA JATC, IEC El Paso, Sheet Metal JATC, UA Local 412, Iron Workers 263, IUEC 131, EPAC and others — so this figure is a floor, not the whole pipeline.
• The apprenticeship denominator is UNMEASURED AT THE RESPONSIBLE AUTHORITY. Asked for capacity during a publicly announced 1 GW buildout, the regional workforce board produced a list it says was 'compiled using publicly available sources' and directed the requester to DOL and TWC, noting those 'may not show the number of current openings'. The pipeline's true size is not merely uncounted here; it is uncounted by the body responsible for planning it.
• Completers are counted at first major only, across all award levels.
• The demand denominator spreads peak need over 3 years, which is generous to the local pipeline.

P4 — Local contractors are not positioned to capture the work

Verdict: CONFIRMED

Every one of the 22 permits produced names an out-of-market firm as licensed professional of record — 6 firms, none headquartered in El Paso. Local share 0%, against a 20% threshold. Both electrical contractors of record are out of market, in the trade the audit identifies as the binding constraint.

Pre-registered rule: CONFIRMED if local share of awards is below 20%, measured on licensed professional of record on issued City permits.

Evidence

• 22 City permits produced for the campus. Local share of licensed professional of record: 0/22 = 0%. _{City of El Paso permit records, Wurldwide LLC / Project Seafox}
• J. E. Dunn Construction (general, Kansas City, MO) — 9 permit(s); not El Paso-headquartered. _{permit record}
• Hensel Phelps Construction (general, Greeley, CO) — 5 permit(s); not El Paso-headquartered. _{permit record}
• Rosendin Electric (electrical, San Jose, CA) — 5 permit(s); not El Paso-headquartered. _{permit record}
• M. A. Mortenson (general, Minneapolis, MN) — 1 permit(s); not El Paso-headquartered. _{permit record}
• Enchanted Rock (general, Houston, TX) — 1 permit(s); not El Paso-headquartered. _{permit record}
• CEC Facilities Group (electrical, Dallas, TX) — 1 permit(s); not El Paso-headquartered. _{permit record}

Caveats

• Measured on licensed professional of record, a conservative proxy for prime awards: it captures who is legally responsible, not every subcontract.
• SUBCONTRACTOR ROSTERS WERE NOT PRODUCED. Local participation at the sub tier remains unmeasured, and local firms may well hold sub work.
• Permit records are OCR'd from scans; the firm names were spot-verified against the source images.

P5 — There is no binding local-hire or local-contract instrument

Verdict: CONFIRMED

1 of 6 binding elements present; 1 carry an enforceable penalty.

Pre-registered rule: CONFIRMED if fewer than 3 of 6 elements present, or none carries a penalty.

Evidence

• local hire percentage: absent — No local hire percentage and no defined geographic hiring area appears in any produced document. The Stan Roberts Development Agreement contains zero instances of 'local hire', 'local labor', 'local business' or 'hiring'. _{City of El Paso council presentation, 2023-12-04 (Wurldwide LLC / Meta, Items 1 & 2); Stan Roberts Development Agreement}
• apprenticeship utilization: absent — Zero instances of 'apprentice' in any produced document. _{City of El Paso council presentation, 2023-12-04 (Wurldwide LLC / Meta, Items 1 & 2); Stan Roberts Development Agreement}
• wage floor: absent — No wage floor or prevailing wage requirement. An $86,500 average salary appears ONLY as an assumption inside the economic impact model, explicitly labelled '*Direct jobs assume an average annual salary of $86,500'. An assumption in a model is not a contractual floor. _{City of El Paso council presentation, 2023-12-04 (Wurldwide LLC / Meta, Items 1 & 2); Stan Roberts Development Agreement}
• reporting cadence: absent — No workforce reporting duty, frequency, format or responsible filer appears in the produced records. _{City of El Paso council presentation, 2023-12-04 (Wurldwide LLC / Meta, Items 1 & 2); Stan Roberts Development Agreement}
• third party verification: absent — No independent verifier named. Separately, the regional workforce board that would be the natural verifier is a Meta donation recipient and marketing counterparty as of August 2026 — see the WSB production analysis. _{City of El Paso council presentation, 2023-12-04 (Wurldwide LLC / Meta, Items 1 & 2); Stan Roberts Development Agreement}
• clawback or penalty: present, with enforceable penalty — A clawback EXISTS but is not tied to employment. Council presentation: 'City may repurchase the land if minimum infrastructure improvements are not made within 5 years of closing.' It is triggered by the $5M infrastructure obligation, not by jobs, local hire, or wages. The only contractual employment term is 'Minimum of 50 FTEs following completion' — with no stated consequence for missing it. _{City of El Paso council presentation, 2023-12-04 (Wurldwide LLC / Meta, Items 1 & 2); Stan Roberts Development Agreement}

Caveats

• Aspirational language without a penalty counts as absent.

P6 — El Paso is tracking Abilene's documented failure path

Verdict: CONFIRMED

El Paso enters its ramp with electrical trade intensity 0.61x Abilene's pre-Stargate level — no better prepared than the metro whose labour pool was documented as depleted.

Pre-registered rule: CONFIRMED if El Paso enters its ramp with trade employment intensity at or below (ratio <= 1.0) Abilene's intensity before the Stargate ramp — i.e. no better prepared than the metro where the failure is documented.

Evidence

• El Paso County, TX (2025Q4): electrical trade employment is 0.83% of all private employment. _{2025Q4 · NAICS 23821}
• Taylor County, TX (Abilene — Stargate) pre-ramp (2023Q4): 1.36% of all private employment. _{2023Q4 · NAICS 23821}
• El Paso's trade intensity is 0.61x Abilene's pre-ramp intensity. _derived

Caveats

• Trade intensity normalises for metro size but not for the size of the buildout relative to the metro. Abilene absorbed a very large project into a much smaller labour market.
• Comparison is directional evidence about preparedness, not a forecast.

Control case

PASS — Control distortion detected. Taylor County, TX (Abilene — Stargate) NAICS 23821 since 2024: peak average weekly wage +18.4%, peak employment +2.2%.

The instrument is checked against Abilene (Taylor County), which received a comparable buildout 18–24 months earlier with documented outcomes: local electrician pool depleted, out-of-state labour imported, Texas license reciprocity opened with Iowa/Alabama/Arkansas, site wages roughly double local subcontractor scale. An instrument that cannot detect a distortion already in the record cannot be trusted on one that is not.

Collection issues

148 issues recorded — suppressed cells and unpublished periods. Listed in full in the findings JSON.

• QCEW not published: Bexar County, TX (San Antonio) 2026Q1
• QCEW not published: Bexar County, TX (San Antonio) 2026Q2
• QCEW not published: Bexar County, TX (San Antonio) 2026Q3
• QCEW not published: Bexar County, TX (San Antonio) 2026Q4
• QCEW not published: El Paso County, TX 2026Q1
• QCEW not published: El Paso County, TX 2026Q2
• QCEW not published: El Paso County, TX 2026Q3
• QCEW not published: El Paso County, TX 2026Q4
• QCEW not published: Taylor County, TX (Abilene — Stargate) 2026Q1
• QCEW not published: Taylor County, TX (Abilene — Stargate) 2026Q2
• QCEW not published: Taylor County, TX (Abilene — Stargate) 2026Q3
• QCEW not published: Taylor County, TX (Abilene — Stargate) 2026Q4
• …and 136 more

How to check this

Every figure carries its source URL, series identifier, and retrieval timestamp in data/raw/_manifest.json and data/normalized/qcew.json. Re-run python3 run_pipeline.py --offline to reproduce this report from the cached bytes; the findings checksum must match.

Generated by the El Paso Data Center Readiness Audit pipeline.

Analysis — Workforce Solutions Borderplex production, DCR-1

Request: DCR-1, "Contracting and award records for the Meta/BlackRock data center campus"
Sent: 2026-08-13, 15:58 MT, from infinitivestorytelling@gmail.com to OpenRecords@BorderplexJobs.com
Received: 2026-09-03. 46 files, 9.8 MB. No cover letter, no exemption log, no redaction notice.
Status of the request: misdirected — the letter is addressed to the City of El Paso and asks for City-held records. Superseded by DCR-1C, filed with the City under W211728-081326.

Correction to the prior assessment

The audit predicted WSB would return "no responsive records," on the reasoning that a workforce board does not hold contracting records. That was wrong. WSB produced 46 files. The prediction was right about the category — no contractor list, no awards, no permits, no Chapter 380 agreement appears anywhere in the production — but wrong about the volume and, more importantly, wrong about the value. The misdirected request reached an entity that turns out to sit close to the centre of Meta's El Paso workforce coordination, and it produced material no City request would have surfaced.

What is NOT in the production

Nothing responsive to DCR-1 items 1–4. A full-text sweep of all 33 extractable documents returns a single hit across the terms general contractor · subcontractor · prime contractor · notice of award · Chapter 380 · local hire · prevailing wage · apprenticeship utilisation · community benefit.

This confirms the City is the correct custodian and that DCR-1C / DCR-2C carry the enforceability questions. P4 and P5 remain INSUFFICIENT.

Finding A — the apprenticeship question is answered, and the answer changes the finding

The audit's standing caveat on P3 was that 32 IPEDS completers a year is a floor, because IPEDS excludes union and merit-shop apprenticeships. The production resolves that caveat in two parts.

Part one — the programmes exist, and there are more than the state list shows.

The state ETPL extract WSB supplied (TWC ETPS Apprenticeship list 08-17-2026, 240 rows, statewide) contains exactly two El Paso entries in the building trades, both CIP 460302:

WSB's own compiled memo (EP County area - Apprenticeship Program) lists ten programmes or sponsors, adding Sheet Metal JATC (El Paso / Southern NM), UA Local 412 (plumber, pipefitter, steamfitter, HVAC), Iron Workers Local 263, IUEC Local 131 elevator constructors, the El Paso Association of Contractors, a Bricklayers JAC, and historically registered programmes at AGC El Paso and the Southwest Specialty Contractors Association — the last three flagged by WSB itself as needing status confirmation.

So the trades pipeline is broader than IPEDS alone shows. The caveat was correctly stated and is now closed on the question of existence.

Part two — nobody holds the capacity numbers.

WSB supplied no enrolment, capacity, or completion figures for any of the ten. Its memo states the list "was compiled using publicly available sources and is provided for informational purposes," directs the requester to DOL's apprenticeship.gov and TWC, and notes those resources "may not show the number of current openings within a specific programme."

This is the finding. The regional workforce development board, asked during a publicly announced 1 GW buildout how much apprenticeship capacity the region has, produced a list assembled from public web searches and no headcount. The problem is not that the pipeline is small — that is arguable. It is that the body responsible for regional workforce planning cannot say how large it is.

Effect on P3: the verdict does not change, but its basis should. The volume test rested on IPEDS alone; it should now be reported alongside the fact that the apprenticeship denominator is unmeasured at the responsible authority. Recommend widening the P3 caveat rather than narrowing it.

Finding B — the independent verifier is becoming a paid counterparty

Two arrangements between Meta and WSB are documented in the production.

A donation, in processing. On 2026-08-07 a Meta community development specialist wrote that Meta had "initiated Meta's supplier onboarding for WORKFORCE SOLUTIONS BORDERPLEX, INC." in order to "enable donated processing," requesting tax identification, company registration and bank account details for payments. WSB completed enrolment on 2026-08-10. Meta's reply the same morning: "next will be contracting." The donation concerns America's Workforce Academy, Meta's skilled-trades training programme.

A marketing agreement, under negotiation. On 2026-08-03 WSB's reviewer circulated requested revisions to a Meta marketing agreement covering case studies, testimonials, quotes and logo use.

WSB's legal review was careful and appropriate, and should be credited as such. The requested revisions moved governing law from California to Texas with venue in El Paso County; removed "irrevocably" from the licence grant; required prior written approval for each individual marketing use; preserved governmental and sovereign immunity; expressly acknowledged that WSB is subject to the Texas Public Information Act; and added a No Endorsement clause stating that nothing in the agreement construes WSB as endorsing "Meta, its products, services, policies, or business practices." That is a public body protecting itself properly, and nothing here suggests otherwise.

The finding is structural, not behavioural. Element 5 of the binding-language checklist requires "a verifier independent of both the City and the recipient." The entity best positioned to verify local workforce outcomes on this project is concurrently a donation recipient from, and a marketing counterparty to, the company whose outcomes require verification. No misconduct is alleged and none is evident. The point is that independence is being consumed by ordinary, well-intentioned commercial arrangements, and that is exactly what a binding verification clause exists to prevent.

Effect on P5: does not resolve it — the agreement text is still not public — but supplies concrete, sourced evidence for why element 5 matters here specifically.

Finding C — two job numbers are in circulation and they differ by twenty times

WSB commissioned IMPLAN modelling of the $10B investment (Meta Implan summaries 3x, 2026-05-02):

Meta's own public figure is 4,000+ construction workers at peak. The IMPLAN figure is roughly twenty times larger.

Both can be technically correct: IMPLAN "jobs supported" counts direct, indirect and induced job-years across the full multiplier chain, not simultaneous headcount, and the document does caveat that these "represent construction-phase impacts and are not permanent positions." But 82,013 is the kind of number that escapes its footnote. If it enters public debate unqualified against a project whose own developer says 4,000, it will mislead badly.

WSB's own analysis reaches the audit's conclusion. The same document states the project "require[s] targeted workforce strategies focused on skilled trades, infrastructure readiness, and specialized technical talent rather than broad-based" job creation, and warns that "large-scale investments may not fully engage local workforce systems without early alignment." That is corroboration of the audit's thesis from inside the body being examined, and it is the most useful single sentence in the production.

Finding D — the contractor orbit is out of market

On 2026-08-13 WSB's CEO wrote that Meta had green-lit two workforce sessions for 25–26 August, and that Meta "invited their partners to attend these sessions with us; Associated Builders & Contractors (Houston), CBRE, and the local prime contractor."

Three things follow. ABC's participating chapter is Houston, not El Paso — and the state ETPL confirms ABC's trades apprenticeship programmes sit in Austin and Corpus Christi, with no El Paso equivalent listed. A local prime contractor exists but is unnamed in the production. And separately, Meta invited WSB to join one of its general contractor meetings in June 2026.

This is a lead, not a finding. But the shape — out-of-market contractor associations arriving with the project while local capacity goes uncounted — is the Abilene pattern, and the local prime's name is now a known unknown worth one targeted question.

Recommended next actions

1. Name the local prime. One short follow-up to WSB, or the attendee list for the 25–26 August sessions, likely produces it. It is the single highest-value missing fact for P4.
2. Re-aim the apprenticeship question. WSB does not hold capacity data; DOL RAPIDS and the TWC 2026 apprenticeship report do. Request those directly rather than re-asking WSB.
3. Widen the P3 caveat in the pipeline to state that the apprenticeship denominator is unmeasured at the responsible authority.
4. Hold P4 and P5. Nothing here resolves them. W211728-081326 / W211732-081326, due 2026-08-27, remain the operative requests.
5. Handle Finding B carefully in print. The documented facts are strong and the framing must stay structural. WSB's counsel did the right things; the story is about what independence requires, not about anyone's conduct.

The Audit That Never Happened

El Paso's own auditors promised to check whether a $2.7 million failure had been fixed. They put it in writing twice. Two years later, not one hour of work has been recorded.

On April 17, 2024, the City of El Paso's Internal Audit Department published a report on how the city hands out economic development incentives. It was not a gentle document.

Auditors pulled seven files for businesses receiving Chapter 380 incentive agreements — the deals where the city forgives taxes and fees in exchange for promised investment and jobs. In six of the seven, there was no documentation that anyone had checked whether the company could actually complete the project it was being paid to build. That is an 85.7 percent failure rate on the most basic question in the program.

They found that the incentive totals reported to City Council left out fee waivers and tax-district reimbursements. Across five agreements, Council was told the city had paid out $5,344,991. The real figure was $6,566,958. Council was working from numbers that understated the cost by roughly 22 percent.

They found that of 115 closed agreements, four had received money without finishing what they promised. Two of those agreements had been written without a recapture clause — no mechanism to claw the money back. One company withdrew. One defaulted. Of approximately $2.77 million at stake, the city recovered $1,043.33.

That is four hundredths of one percent.

Management accepted every finding. The Economic and International Development Department, then under Interim Director Karina Brasgalla, committed to fixing all five: a formal financial-analysis process, complete incentive reporting to Council, deadline tracking, closure verification. The implementation date written into the report was January 1, 2025.

The obvious next question is whether any of it happened. In government auditing, that question has a formal answer, and it is not optional. The city's own Internal Audit Policies and Procedures Manual states it plainly on page 72: federal auditing standards and the Institute of Internal Auditors both "require a post audit follow-up on all audit recommendations made to determine that appropriate action was taken."

So the department scheduled one.

Scheduled twice. Performed never.

The FY2024–2025 Annual Audit Plan, signed by then-Chief Internal Auditor Edmundo Calderon on September 1, 2024, lists it in the fourth quarter, June through August 2025: "Follow Up Audit: Economic Development – 380 Agreement Monitoring Audit."

It was not done.

The following year's plan, issued September 1, 2025, schedules it again — second quarter, December 2025 through February 2026, with 250 hours budgeted. But this version carries a qualifier the first one didn't have. In Attachment 6, next to the follow-up, the plan reads: "Dependent on FOAC decision."

Then comes the number that ends the argument. The department publishes quarterly updates showing hours actually worked against every scheduled engagement. The first-quarter update, signed by Deputy Chief Internal Auditor Elizabeth De La O and dated November 30, 2025, shows several second-quarter engagements already underway — the police overtime audit with 96.5 hours logged, an environmental services review with 114.5, a housing renovation audit with 31.25.

The 380 follow-up shows zero.

It is the only second-quarter engagement with nothing charged to it at all. And as of the department's most recently published reports, dated May 18, 2026, no follow-up report exists.

More than two years after the audit. Seventeen months past the date the fixes were due. Two annual plans, both approved by the Financial Oversight and Audit Committee, both promising the same verification. Not one recorded hour.

Nobody has said the problems were fixed. Nobody has said they weren't. The city simply stopped looking, twice, and told no one.

What makes that harder to explain is who asked for the audit in the first place.

The 380 review was performed at the request of the Financial Oversight and Audit Committee itself. In July 2024, three months after the report came out, District 4 Representative Joe Molinar — a member of that committee — went on television to describe what it found.

"It shows favoritism to certain businesses doing business with the city of El Paso, and it's not right," Molinar told KFOX14/CBS4. "And I'll say right now, it probably looks like corruption."

He was talking about his own city's incentive program, in public, as a member of the body charged with overseeing it. He also said what would happen next: "Molinar said the FOAC and the auditor's office will work together to make sure that these mistakes won't happen again."

The committee that made that commitment consisted of Chairman Brian Kennedy and Representatives Josh Acevedo, Joe Molinar, and Art Fierro. Kennedy signed the FY2024–2025 audit plan — the one scheduling the follow-up that never happened.

Not one of them sits on the committee today. Kennedy and Molinar left the Council. Fierro is not on the current roster. Acevedo, who chaired the committee when the second plan was approved in September 2025, was replaced in January 2026. The Financial Oversight and Audit Committee is now Chair Alejandra Chávez, Chris Canales, Deanna Maldonado-Rocha, and Ivan Niño, all appointed by Mayor Renard Johnson, with the Chief Internal Auditor and City Manager Dionne Mack sitting as non-voting members.

This is how an unkept promise survives. Every person who ordered the audit, heard the findings, and pledged to make sure it wouldn't happen again has rotated off the body that made the pledge. The commitment stayed on the audit plan. The people who made it did not.

The word "corruption" was a representative's characterization, not an audit finding; the report itself alleges no wrongdoing and names no scheme. But a sitting oversight official said it on the record, promised the committee would follow through, and the follow-through was scheduled twice and performed zero times.

What else never made the plan

If the follow-up were an isolated slip, it would be a management failure and not much more. It isn't isolated.

In December 2023, City Council approved an incentive agreement with Meta for a data center in northeast El Paso. Meta's stated commitment has since grown from $800 million to $10 billion. El Paso Electric describes the site scaling toward roughly one gigawatt of electrical demand — a load comparable to a small city. The company is projected to become the largest property taxpayer in El Paso — but not the way that phrase sounds.

The December 2023 agreement grants Meta an 80 percent break on city taxes, reported as running 25 years, though accounts of the phased structure vary, with some describing terms extending to 35 years across all phases. Meta pays the remaining 20 percent. The city estimates that share alone at roughly $15 million a year — still more than any other taxpayer in El Paso. The city separately committed $12.5 million for road improvements near the site.

The performance thresholds written into the deal are $800 million in investment and 50 jobs. Meta has since announced $10 billion and 300 workers, and the abatement applies to the full amount. Which means the binding obligations the city can actually enforce are the small ones, and the enormous figures in the press releases are not what the contract requires.

Verifying that a company met its investment and job commitments, and reporting the full cost of an incentive package to Council — including road money and fee waivers, not just the tax line — is precisely what the 2024 audit found the city was failing to do. On the largest such agreement in El Paso's history, no one has checked.

The FY2024–2025 audit plan was issued nine months after Council approved that deal. Its Attachment 6 lists the fourteen new engagements scheduled for the year: peer-review self-assessments, budget transfers, park permits, procurement card reviews, vehicle allowances, a public health cash count, Davis-Bacon compliance, a zoo membership review, an animal services construction audit, a grant program review, executive leave balances, facility maintenance, travel approvals.

None of them touches the Meta agreement. None touches the 380 program. None touches Tax Increment Reinvestment Zone #10, or any other individual development deal.

This is not because the auditors lack a category for it. The city's own risk assessment scores "380 Agreements" as an auditable area — in the current year's plan it carries a weighted risk score of 38, among the highest of any activity in city government, driven by maximum ratings for budget risk, strategic risk, and compliance risk. The department scores the risk. It publishes the score. Then it schedules other things.

Part of the reason is structural. The audit universe is built from the city's organizational chart — departments, divisions, offices. A transaction, no matter how large, has no line to appear on. A billion-dollar agreement is nobody's department.

The city commissioned this, sat for it, and adopted it

In October 2024, City Council hired the accounting firm Weaver and Tidwell to assess whether the Internal Audit Department was doing its job, at a cost reported at $128,000. Weaver interviewed the mayor, every council member, the audit committee, city management and the audit staff on February 6 and 7, 2025, then came back to validate what it had found with the same people. The preliminary results were presented to the audit committee that June.

The verdict: the department is operating at the "Repeatable" level — level two on a five-point scale, with some progress into level three. Foundational practices exist. Consistency does not.

Two of the findings in the completed assessment read like they were written for this article.

On planning: "It is not clear how current audit planning and risk assessment results are fully aligned, which may limit the Internal Audit function's ability to address the most critical areas." The firm also recorded "limited visibility into how audit objectives are selected and prioritized."

On the committee that oversees the office: "Oversight activities are reactive and personnel-dependent" — FOAC oversight "is largely driven by the chair," with "no formalization of FOAC's oversight responsibilities." Among the things Weaver listed as missing: "written procedures documenting oversight practices for continuity across leadership changes."

That last one is the whole problem stated in a single line by the people the city hired to state it. A control that depends on a particular seat being occupied is not a control. It is a habit, and habits do not survive a vacancy.

The assessment carrying those findings is dated September 2025 — twenty months after the 380 fixes came due, and before the follow-up was rescheduled and skipped a second time.

It is worth being precise about what that misalignment is and isn't. The tempting conclusion is that the audit department shields economic development. The department's own numbers do not support it. Economic and International Development received one audit and one follow-up over five years — more coverage than Animal Services, Community and Human Development, Purchasing, Code Enforcement, Real Estate, or Information Security Assurance, all of which score at or above it on the city's risk scale and all of which received nothing at all in the same period. The audit plan has capacity for roughly eight engagements a year against 186 scored areas. Ninety-six percent of the city goes unexamined annually. The scores are published; what connects them to the schedule is not.

The exception shows up somewhere else — not in what gets audited, but in what gets checked afterward. Across the FY2024–25 and FY2025–26 plans, eleven follow-up audits were scheduled: accounts receivable, cyber security, street permits, military affairs travel, planning permits, on-call services, park usage permits, airport accounts payable. Nine were completed or have hours recorded against them.

The two that weren't are the same engagement, scheduled twice. It is also the only follow-up in either plan with a condition attached to it.

The final assessment, dated September 2025, says the same thing. It runs 70 slides, assesses 14 component areas, and lays out 45 specific action items with three-to-twelve-month and twelve-month-plus horizons and a priority ranking. Its verdict on maturity is word for word the preliminary's: Repeatable, with some progress toward Defined.

Its section on follow-up is the one that matters here: "Follow-up audits lack clear linkage to original audits, management accountability regarding responses and action plans is unclear, and disagreements on corrective actions are addressed ad-hoc."

That last clause is worth holding onto. The 380 follow-up carried a qualifier — "Dependent on FOAC decision" — and the most plausible reading of it is a disagreement about scope: follow up on the recommendations, or audit the program again. Weaver's finding is that exactly this kind of disagreement gets handled case by case, with no process behind it.

And City Council adopted the assessment. It went to a Council Work Session on November 17, 2025, as agenda item 25-1366, where the recorded action was "Adopted" and the result was "Pass." Weaver's engagement director, Holly Hart, told Council that advancing the function admits no negotiable items, and that the foundation of the roadmap is its governance section — the part addressed to the committee and the Council themselves.

So the city has not been left guessing about what to fix. It commissioned the assessment, its committee members sat for the interviews in February 2025, its consultant returned to validate the results with them, and its Council adopted the whole thing by recorded vote. The department now has 45 accepted action items of its own, which is precisely the position Economic Development was in after the 2024 audit. Whether anyone tracks and closes them is the same question this article started with, pointed at the office that asks it of everyone else.

Both versions were presented in public — the preliminary to the audit committee on June 17, 2025 and to Council on June 23, the final to Council that November. The final is posted on the city's agenda portal as an attachment to the November meeting. Neither appears on the Internal Audit Department's quality assurance page, which lists external reviews of this office going back to 2007. A resident looking for it where the city files every other outside review of its auditors will not find it, and cannot find it without already knowing which meeting to search.

An office under pressure

None of this happened in a vacuum.

In 2023, Chief Internal Auditor Edmundo Calderon notified the city that he intended to file a whistleblower claim, and in December of that year he sued. He alleged that city officials harassed and intimidated him after his office audited city fuel cards — an audit that began with an anonymous employee tip that a council member's city-issued fuel card was being used to fill the vehicles of campaign workers. The suit named former City Manager Tommy Gonzalez, City Attorney Karla Nieman, and others. The allegations were never tested in court. The city settled in September 2024.

In June 2025, Council met behind closed doors to review the city manager, the city attorney, and the auditor. Weeks after the Weaver assessment became public, Calderon announced his retirement. His last day was August 1, 2025, after twenty years with the city.

The position sat vacant for seven months.

How the city filled it is documented, because records were requested and produced. There was no executive search firm. The job was posted to six professional job boards — five of them bought in a single afternoon — at a total cost of $454, for a department-head position advertised at $180,000 to $205,000. One of the documents the city produced as proof of advertising turned out to be a screenshot of a different, already-closed listing.

None of that is hard to do better, which is the frustrating part. A defensible search for this particular job is not exotic: the chair of the audit committee runs it rather than Human Resources, because the Charter gives the chair operational oversight of the office. The committee approves a written position profile before anything is posted. The interview panel holds a majority from outside city management — audit committee members plus an outside public-sector auditor, which the Institute of Internal Auditors and the Association of Local Government Auditors both exist to supply. The city either retains a search firm from the roster ICMA publishes for exactly this purpose, or records in the minutes why it chose not to. Scoring criteria are fixed before the first application is read. Licenses and references are verified before an offer goes out, not after. And the committee sends Council a written recommendation stating how many people applied, how many met the minimum qualifications, who sat on the panel, and why this candidate. Council holds the appointment power; it should be exercising it on a record.

None of that protects a candidate from scrutiny. It protects the office from the suspicion that the person auditing city management was chosen by city management. That protection is the entire point of the position, and it costs about the price of doing the paperwork.

City Council appointed Adrian Serrano on March 3, 2026. He has more than thirteen years of internal audit experience, and nothing in the public record reflects on his qualifications or his conduct. He inherited all of this. What follows is not about him — it is about the process that reached him.

Who the auditor actually answers to

In May 2023, El Paso voters amended the City Charter. Section 3.20 B.5 now reads:

"the Chief Internal Auditor shall be appointed and removed by, and report to the Council. The Chairman of the Financial Oversight and Audit Committee shall maintain operational oversight over the internal audit function. The City Manager shall be responsible for the implementation of any audit recommendations for changes to City administrative procedures and operations as requested by the Council."

Voters did that deliberately, and they knew exactly what they were changing.

Until 2023, the Charter let the city manager appoint and remove the internal auditor, subject to Council approval, and gave the manager operational oversight of the auditor's work, findings, and recommendations. Proposition J was put on the ballot by District 1 Representative Brian Kennedy — later the chair of the audit committee, and the man who signed the FY2024–2025 audit plan. His argument was one sentence:

"Currently, if there's an investigation of the city manager or any department that works for him, he is in control of the operational end of that examination."

El Paso voters approved it on May 6, 2023, with about 56 percent in favor. Day-to-day oversight moved to the chair of the audit committee. The City Manager's remaining role is implementing audit recommendations when Council asks. The city's own budget book prints the new reporting line as a single vertical column: City Council, then Chief Internal Auditor, then staff. No manager appears in it.

Which is what makes the record of the 2025 search matter. There was no search firm. The postings were bought by staff. The listed contact was inside the audit department itself. And the audit committee — the body voters had just handed operational oversight — appears nowhere in the documents the city produced.

That is not simply a thin search. It closely resembles the arrangement El Paso voters had abolished two years earlier: the hiring of the auditor running through the management structure the auditor is supposed to examine. Council cast the appointing vote, so the Charter's formal requirement was met. Whether the process that produced the candidate followed the structure voters adopted, or reverted to the one they rejected, is answerable only from the selection records — which have been requested and not produced.

Council did vote on the appointment, so the formal requirement was met. But the records produced show no participation by the Financial Oversight and Audit Committee or its chair in advertising, screening, or recommending candidates for the office they are charged with overseeing. The listed point of contact for the job postings was the Deputy Chief Internal Auditor — a position inside the department, below the vacancy being filled.

There is no evidence in the record that anyone did anything improper. There is also no evidence that the oversight body the voters designated had any hand in choosing the officer it oversees. What Council received before it voted, and from whom, is not in any public document.

The hotline

Buried in the Weaver assessment is a finding that deserves more attention than it has received.

The city's employee hotline — the channel through which staff report fraud, waste, and abuse — is run by the Chief Internal Auditor's office. Weaver found that the hotline's procedures manual lacks provisions for independent review of the program's effectiveness, lacks secondary review mechanisms, and lacks defined anti-retaliation controls, including any definition of retaliation or associated training.

And this:

"Currently, the CAE is solely responsible for receiving reports and determining whether they should be escalated and/or closed."

One person decides which reports go forward and which ones close. There is no second set of eyes, no periodic review of whether the system works, and no written protection against retaliation for the employee who calls it.

Remember where the fuel card audit started: an anonymous employee tip. Remember how it ended: with the city's chief auditor alleging he was intimidated, and the city writing a settlement check.

What this costs, right now

At a town hall in August 2026, hosted by City Representative Cynthia Boyar Trejo, residents packed a room past capacity — some listening from the doorway — to ask about Meta's water and power use. One asked officials to produce the scientific studies behind their assurances.

El Paso Water's president, John Balliew, said the utility will hold Meta to a contractual maximum of 2.5 million gallons a day and that the meter physically cannot pass more. He said the demand is "exactly the amount of water that we were planning to provide for customers in that area anyway." He also acknowledged that the plant receiving the site's wastewater cannot remove high salt concentrations, and that discharge terms are still being negotiated — while the site already has construction power and is expected to begin operating this year.

The utility has compared Meta's water use to a golf course. The comparison is Painted Dunes, in northeast El Paso. Painted Dunes runs on reclaimed water. Meta will draw from the potable system, because El Paso Water says its available reclaimed supply is already committed.

El Paso Electric's Daniel Perez was asked whether the project could raise residential rates. "Absolutely not," he said, citing the regulatory structure.

Six weeks before he said that, on June 23, City Council had voted unanimously to negotiate a community benefits agreement with Meta. Among the provisions Representative Cynthia Boyar Trejo laid out was a "Ratepayer Protection Fund" — money Meta would put up, distributed by the city, "to help El Paso Electric customers with any bill increases as a result of the data center's operation." The same item sought funding to strengthen the solar and battery grid to reduce brownouts, and infrastructure money to leave more water in the local aquifer for residents.

A city does not negotiate a fund to cushion bill increases from a project that cannot cause bill increases. Either the fund is unnecessary or the assurance is wrong. Eleven days before the town hall, the city had also asked the state to extend data-center ratepayer protections to El Paso, and Council had voted to intervene at the Public Utility Commission in El Paso Electric's proposed large-load tariff, hiring outside experts to make sure households don't subsidize the data center. Those are not the actions of a government that believes the answer is "absolutely not."

There is a narrower reading available to the utility — that Perez was speaking about regulated residential rates, while the fund addresses bills, which move with usage and other costs. That distinction may hold. Testing it requires reading the tariff's cost-allocation provisions against the assurance, which is verification work, and verification work is the thing this city has not been doing.

It is also worth knowing who is at the table. The community benefits agreement is to be negotiated through roundtables seating El Paso Water, El Paso Electric, the El Paso Chamber, the Greater El Paso Association of REALTORS, an environmental advocate, Meta — and one resident representative. The project itself was recruited by the Borderplex Alliance, the regional economic development organization, whose executive committee includes El Paso Electric's chief executive. None of that is improper, and in a region this size some overlap is unavoidable. It does mean that most of the voices assuring El Pasoans about this project have an institutional stake in it proceeding, which is precisely the condition that makes independent verification valuable rather than optional.

Every one of those statements is checkable. There is a contract, a meter specification, a set of planning documents, a tariff filing in a public docket. Verifying claims like these against the underlying records is precisely what an internal audit function exists to do.

There is no engagement on any city audit plan that tests a single one of them.

Three things to watch

September 1. The FY2026–2027 Annual Audit Plan is due. It is the first document the new Chief Internal Auditor puts his name to, and the clearest test available. Three questions answer themselves on reading it: Does the 380 follow-up finally appear, and without a discretionary condition attached? Does the audit universe include major development agreements as an auditable area, so that transactions like Meta have a line to appear on? Are the risk scores published alongside the engagement list, so anyone can see whether the highest-scoring areas are the ones getting audited?

A plan that scores 380 Agreements at 38 and schedules nothing, again, without explanation, is an answer.

The missing records. Four documents would settle most of the open questions, and any member of the public can request them under the Texas Public Information Act. The Weaver and Tidwell engagement contract and any deliverables beyond the assessment itself. The minutes of the Financial Oversight and Audit Committee's meetings of October 9, 2025 — where the 380 audit results were an agenda item — and November 14, 2025, where the Weaver assessment was discussed. The notifications the city's own manual requires management to send the auditor when corrective actions run past their deadline; the 380 fixes were due January 1, 2025. And native-format copies of the annual audit plans, which the city currently publishes as scanned images that cannot be searched or read by screen readers.

The committee. The Financial Oversight and Audit Committee — Chair Alejandra Chávez, Chris Canales, Deanna Maldonado-Rocha, and Ivan Niño, with terms running to January 2027 — holds the charter duty of legislative oversight over this function. The Chief Internal Auditor reports to the chair. Its published duties include monitoring the implementation of all audit recommendations and presenting peer review results to Council annually. The last external peer review of the department was published in 2022; the department's own manual says these should occur every three years. A review budgeted 500 hours in the current year shows zero hours worked.

The committee meets quarterly and takes public comment. Three questions belong on that record: Why was the 380 follow-up scheduled twice and performed neither time, and who made the decision it was "dependent" on? Will the assessment the Council adopted be posted on the quality assurance page, where every other outside review of this office is filed? When will the overdue peer review be completed?

Every fact above comes from a City of El Paso document — audit reports, annual audit plans, quarterly plan updates, the Policies and Procedures Manual, the City Charter, the budget book, and records produced under public information requests — or from published reporting, identified as such. Where a figure could not be traced to a document, it was left out. The full technical version of this analysis, including the scored standard and complete source table, is available as a companion document.

Where the Audit Function Actually Goes

A capacity and coverage analysis of the City of El Paso Internal Audit Department, FY2021–FY2026

Date: August 11, 2026
Sources: FY2025–2026 Annual Audit Plan, Attachments 1, 2, 5, and 6; FY2024–2025 Annual Audit Plan; FY2026 Q1 Plan Update; FY2025 Proposed Budget Book. All figures reconcile to the totals printed in those documents.
Companion to: External_Auditor_Roadmap_El_Paso.md §6a

1. The question

Two claims were tested against the department's published data:

1. The observation: Economic & International Development received one audit and one follow-up in five years — more coverage than several areas scoring higher on the city's own risk scale that received nothing at all. What does that imply system-wide?
2. The proposed inference: "By focusing on economic development, other parts of the city government are being neglected."

Verdict on claim 2: false as stated — and the true explanation is more serious.

2. Economic development is not crowding anything out

Across FY2021–FY2025 the department completed 114 engagements: 32 audits, 18 follow-ups, and 64 projects.

Economic & International Development received 2 of 114 — 1.8% of all audit activity in five years.

A function consuming 1.8% of capacity cannot be crowding out the other 98.2%. Economic development is not over-audited relative to anything. It is barely audited, alongside almost everything else.

For scale: the Tax Office received 21 engagements — 10.5 times as many — and every one of them is a recurring "project," not an audit. The department's single largest area of activity over five years is a quarterly refund review.

3. Where the capacity actually goes

FY2025–26 budgeted 12,782 audit hours. Allocation, reconciled exactly to Attachment 1:

Nearly half the department's year is committed before a single risk-based decision is made.

The recurring block, ranked

Two observations that carry the section.

First, supporting the FOAC consumes 2,000 hours a year — roughly one full-time position, and 15.6% of all audit capacity. That is eight times the 250 hours budgeted for the 380 follow-up, and four times the largest single named engagement in the plan. The committee that decided the 380 follow-up was "dependent on FOAC decision" is itself the department's single largest consumer of audit hours.

Second, strip out FOAC support and what remains of the recurring block is revenue assurance: sales tax, franchise fees, hotel occupancy tax, tax refunds, sales tax discovery. These protect money coming in. Almost none of the recurring capacity is directed at verifying money going out — which is precisely the category 380 agreements fall into.

The department functions, structurally, as a revenue-assurance shop with a risk-based audit function attached to whatever hours are left.

4. Where the engagements go

Of 50 auditable areas listed in the five-year table:

• 30 — 60% — received zero assurance work (no audit, no follow-up) in five years.
• 22 received no engagement of any kind, not even a project.
• The top five departments captured 54% of all activity.

(Conservative count: six organizational header rows for Deputy City Manager portfolios were excluded. Including them raises the zero-assurance share.)

The zero-assurance list, against the city's own risk scores

Two of these deserve specific notice.

Housing and Grant Administration (41 and 40) administer federal funds. Five years without a single internal audit engagement is a material coverage gap in the area of city operations carrying the most external compliance exposure. (The city receives an external Single Audit under Uniform Guidance; this finding concerns internal audit coverage, which is a different and complementary control.)

Information Security Assurance — Governance, Risk Management, and Controls, each scored 38 — has received no engagement in five years, in a city that has just onboarded a gigawatt-scale hyperscale data center and whose FY2025–26 plan devotes 500 hours to "Cyber Audits (Administration)" without naming an engagement.

5. The risk assessment is decorative

The department scores 186 auditable areas annually across nine weighted criteria. Attachment 6 of the FY2025–26 plan states the year "had the capacity for eight (8) engagements."

Eight slots against 186 scored areas is roughly 4% annual coverage. At that rate, cycling the audit universe once takes about 23 years — longer than most municipal careers, and far longer than the 25–35 year term of the Meta abatement.

This is the mechanism behind the finding in §6a of the roadmap that risk score does not predict audit selection. It cannot. When 96% of the universe goes unexamined every year, a score of 41 and a score of 20 produce the same outcome for almost everyone, and the ranking carries no operational consequence. The city performs the analytic ritual of risk-based auditing and then allocates by other means — carry-forwards, recurring obligations, management requests, and committee direction.

6. The true version of the claim

The proposed inference — that focusing on economic development starves the rest of the city — is not supported. But there is a defensible version, and it is worse:

The city expanded its risk surface enormously without expanding the function that verifies it.

Between FY2024 and FY2025 the Internal Audit Department's authorized staffing was flat at 10 positions. By the FY2025–26 plan the department was operating with 8 filled positions and 2 vacancies — including the Chief Internal Auditor, a seat that stayed empty for roughly seven months. Over the same period the city approved a data-center agreement whose announced value grew from $800 million to $10 billion, expanded TIRZ #10, and entered negotiations on a community benefits agreement with utility, water, workforce, and environmental provisions.

Risk grew by an order of magnitude. Assurance capacity did not move, and in practice declined.

So the correct statement is not that economic development crowded out other audits. It is that economic development crowded out nothing — because it was never given room in the first place — while simultaneously generating the largest new risk exposure in the city's history. Both the neglected departments and the unverified incentive agreements are symptoms of the same undersized, obligation-bound function.

7. Implications

For the El Paso argument. The failure to verify the 380 program is not evidence of favoritism toward economic development. It is evidence of a function with ~4% annual coverage, a third of its hours pre-committed to recurring work, 15.6% consumed by supporting the very committee that deferred the follow-up, and no capacity headroom to absorb a new billion-dollar risk. That framing is harder to rebut than a favoritism claim and it points at a fixable cause.

For the instrument. Coverage ratio — scored areas versus annual engagement capacity — should be a scored control in the national study, not merely a recorded covariate. A verification function below a coverage threshold cannot perform risk-based selection regardless of independence, which means capacity and independence interact rather than compete. This refines the H2 rival: capacity may be a necessary condition and independence a sufficient one.

For the public record. The single most quotable pairing in this analysis: 2,000 hours a year supporting the audit committee; 250 hours budgeted and zero hours spent verifying a $2.77 million failure the same committee ordered investigated.

Five Things to Look For in El Paso's Next Audit Plan

Published August 2026, before the plan exists. Scored publicly within seven days of its release, whichever way it goes.

The City of El Paso's Internal Audit Department issues an Annual Audit Plan on or about September 1 each year. The FY2026–2027 plan will be the first bearing the name of Chief Internal Auditor Adrian Serrano, appointed by City Council on March 3, 2026.

I am publishing what I will look for before the document exists, so that the standard cannot be adjusted after the fact to fit whatever arrives. Each test below has a stated pass condition, a stated fail condition, and a source in the city's own records. Within seven days of the plan's publication I will post a scored result against these five tests — including if the plan passes all of them, which I will say as plainly as I would say the opposite.

This is not a prediction that the plan will fail. It is a commitment to judge it against criteria fixed in advance. That is the same discipline the audit profession requires of its own work, and it is the only thing that separates analysis from commentary.

Test 1 — Does the 380 Agreement follow-up appear, without a condition attached?

Background. In April 2024 the city's own auditors reported that 6 of 7 sampled Chapter 380 incentive files contained no documented financial analysis; that incentive costs reported to Council understated the total by roughly 22 percent by omitting fee waivers and TIRZ reimbursements; and that four agreements closed without completion, involving about $2.77 million, of which the city recovered $1,043.33. Corrective actions were due January 1, 2025.

The follow-up audit was scheduled in the FY2024–2025 plan (fourth quarter) and again in the FY2025–2026 plan (second quarter, 250 hours) — the second time with the qualifier "Dependent on FOAC decision." The department's own quarterly update shows zero hours charged as of November 30, 2025. No follow-up report appears among published reports through May 2026.

PASS: the follow-up appears as a scheduled engagement with budgeted hours and no discretionary condition — or the plan, or the record accompanying it, documents a legitimate reason for deferral.

FAIL: it is absent, or present again with an unexplained conditional, with no accompanying record.

The most likely explanation, stated first. Both misses fall inside a leadership gap. The FY2024–25 engagement was scheduled for June–August 2025 — the quarter the Chief Internal Auditor announced his retirement and left. The FY2025–26 plan was issued one month into the resulting vacancy, and its second-quarter window fell entirely within the seven months the office had no chief. The conditional is not about whether to act but about scope — whether to follow up on the recommendations or conduct a new audit of the program — which is exactly the question a department without a chief must refer to its committee and cannot drive to resolution.

That is a mundane and largely exculpatory account, and it should be said before anything harsher. It does not, however, dispose of the matter. Four other second-quarter engagements accrued hours during the same vacancy, so the vacancy did not halt work generally. And if the absence of one officeholder can suspend verification of a $2.77 million control failure for two planning cycles, then the verification was never a process — it was a person. That is a structural finding, not an accusation, and it is the more useful one.

A second sound reason to defer. If management formally extended its corrective-action timeline, auditing implementation before implementation is complete tests nothing. Deferral on that basis is standard practice, not avoidance.

If that is what happened, it should be visible. The department's own Policies and Procedures Manual (p. 72) provides that "if there are delays in implementing corrective action, the Chief Internal Auditor should be notified immediately as to the nature and reason for the delay." An extension therefore generates a record.

So the test is really two questions: does an extension memo or equivalent record exist, and was the deferral disclosed rather than left as an unexplained conditional? A documented extension is a PASS. Three years of scheduling with no extension record and no explanation is a FAIL. That distinction is the whole test, and it is answerable from documents.

Test 2 — Can a major transaction appear on the audit plan at all?

Background, stated precisely. The department builds its audit universe from the city's organizational chart — departments, divisions, offices. "380 Agreements" does exist as a scored line within Economic & International Development, carrying a weighted risk score of 38. What does not exist is any unit of assessment below the program level.

That distinction is the point. A program is scored once, at one score, regardless of whether it contains a $50,000 façade grant or a multi-billion-dollar data-center agreement. There is no mechanism by which an individually material transaction — the Meta agreement, TIRZ #10 — is scoped, scored, or scheduled on its own terms. It is absorbed into a program line and inherits that line's single fate: audited once in five years, or not at all.

This is why the FY2024–25 plan's fourteen engagements included park permits, procurement cards, and a zoo membership review, but nothing touching a transaction Council had approved nine months earlier.

PASS: the plan provides some route by which an individually material agreement can be scoped and scheduled — a major-agreement or high-exposure contract category, a materiality threshold triggering separate assessment, or a named engagement on a specific agreement.

FAIL: program-level scoring remains the finest available resolution, so that the size of a single commitment cannot affect whether it is examined.

Test 3 — Are the risk scores published alongside the engagement list?

Background. The department scores 186 auditable areas annually against nine weighted criteria. "380 Agreements" carries a weighted score of 38 in the current plan, among the highest in city government. The plan states it had capacity for eight engagements — roughly 4 percent annual coverage.

At that ratio the score cannot determine what gets audited, and the plan does not publish the reasoning that connects scores to selections. Nobody outside the department can currently tell what does.

PASS: the risk assessment is published with the plan, scores are legible, and the engagement list is traceable to them.

To be explicit, because this test is easy to misread as demanding mechanical rule-following: professional auditors legitimately override numerical rankings using judgment — management requests, hotline referrals, mandated engagements, emerging risk. Nothing here disputes that. Documented judgment is a PASS. A plan that says "these areas scored highest; we selected differently for these stated reasons" satisfies this test completely.

FAIL: scores omitted; or scores published with no stated basis for selection at all, so that judgment and inertia are indistinguishable from outside; or published in a scanned format that cannot be searched or read by assistive technology, as the FY2024–25 plan was.

Test 4 — Is the overdue external peer review actually performed?

Background. The department's Policies and Procedures Manual states the Financial Oversight and Audit Committee requests external peer reviews every three years. The last published review is dated 2022, making it due in 2025. The FY2025–26 plan scheduled "Peer Review of the Internal Audit Department" for the third quarter with 500 hours budgeted; the quarterly update shows zero charged. A related "Red Book Assessment" drew 15.25 hours against 250 budgeted.

PASS: the review is completed and published, or scheduled with hours and a date certain.

FAIL: carried forward a third year without explanation.

One substitution that would not satisfy this test. The city could reasonably point to the $128,000 Weaver and Tidwell engagement as external scrutiny already performed. It is not a peer review, and Weaver says so itself: the report notes that the function "has demonstrated necessary alignment with relevant audit standards via required peer reviews" — treating the peer review as a separate, existing mechanism its own maturity assessment does not replace. The department's Manual likewise records that it "has chosen to participate in ALGA's peer review program." A maturity assessment and a conformance peer review answer different questions, and only one of them yields a pass/fail opinion on conformance with professional standards.

Test 5 — Does the plan account for the 45 recommendations the city already paid for?

Background. City Council engaged Weaver and Tidwell LLP for $128,000 to assess the audit function. The final report, delivered September 2025, placed the department at the "Repeatable" level — level two of five — and set out 45 action items across 14 component areas, each with an assigned owner under a RACI framework, an implementation horizon of 3–12 months or 12+ months, and a priority rating. Short-term items began expiring in December 2025; the full short-term window closes this September.

The report also found that the department's own "follow-up audits lack clear linkage to original audits" and that "management accountability regarding responses and action plans is unclear."

PASS: the plan allocates hours to tracking, verifying, and reporting progress against the 45 action items — and, for those items Weaver assigned to the Chief Internal Auditor as Responsible, to carrying them out.

FAIL: no reference to the assessment the city bought, in the first plan issued after receiving it.

A note on independence, since it cuts both ways here. Internal auditors must not assume management's responsibilities; implementing operational changes in an audited department would impair objectivity for future audits of it. That principle does not bar the Chief Internal Auditor from improving his own function — Weaver's RACI assigns the CAE as Responsible on many of the 45 items, precisely because they concern internal audit's own methodology, training, and reporting. Running your own department is not assuming management responsibility over an auditee.

Where items are assigned to the City Manager, City Attorney, or department directors, internal audit's role is to monitor and report, not to execute. This test is written to match that split, and a plan that observes it scores full marks.

What this is, and is not

It is not an allegation of wrongdoing. The Weaver assessment examined process maturity, not transactions. Its stated scope was to assess structure, roles, responsibilities, and practices against professional standards — it was not a financial or forensic review, and nothing in it concerns misappropriation. Nothing in this document does either.

It is not about the current Chief Internal Auditor's performance. He inherited every condition described here, and no test above requires him to resolve three years of accumulated backlog in a first annual plan. Each pass condition can be satisfied by scheduling, disclosure, or written explanation — none requires completed remediation. The FY2026–2027 plan is simply the first artifact of his tenure that can be measured, and measuring it is not the same as blaming him for what preceded it.

Where the answer is the committee's, the score follows the committee. Under the City Charter as amended by voters in May 2023, the Chief Internal Auditor is appointed by and reports to City Council, with the Chairman of the Financial Oversight and Audit Committee holding operational oversight of the function. If the FOAC directed a deferral — of the 380 follow-up, the peer review, or anything else — that is the committee's decision to make, and making it is entirely legitimate. A documented committee direction is a pass for the auditor. It also moves the question to the body that gave it, which is where the Charter puts it. What no one can defend is a deferral with no decision-maker attached to it.

Every fact above comes from a City of El Paso document, listed below. Where a figure comes from press reporting rather than a document I have read in full, it is marked as such.

The plan is due September 1. The score follows within seven days.

A standard published in advance can be met. That is the point of publishing it in advance.

Sources

Economic Development – 380 Agreement Monitoring Audit, No. A2023-07 (City of El Paso Internal Audit Department, issued April 17, 2024). Published by the city. → 6 of 7 sampled files without documented financial analysis (Finding 2); incentive reporting to Council understated by $1,221,967 across five agreements, ~22% of the actual total, from omitted fee waivers and TIRZ reimbursements (Finding 3); four agreements closed without completion totaling $2,773,383.33, of which $1,043.33 was recovered (Finding 5); management responses with an implementation date of January 1, 2025.

FY2024–2025 Annual Audit Plan (dated September 1, 2024). Published by the city as a scanned image without extractable text; read via optical character recognition. → Attachment 1, Fourth Quarter: "Follow Up Audit: Economic Development – 380 Agreement Monitoring Audit." Attachment 6: the fourteen new engagements scheduled for the year. Audit universe constructed from the city organizational chart as of 8/31/2024.

FY2025–2026 Annual Audit Plan (dated September 1, 2025). Published by the city. → Attachment 1, Second Quarter: "Follow-Up: 380 Agreement Monitoring Audit," 250 budgeted hours. Attachment 6: the qualifier "Dependent on FOAC decision," and the statement that the year had capacity for eight engagements. Attachment 2: 186 scored auditable areas across nine weighted criteria, with "380 Agreements" at a weighted score of 38. Attachment 5: five-year completion history.

FY2026 First Quarter Audit Plan Update (as of November 30, 2025; signed and distributed to the FOAC, City Manager, and Chief Financial Officer). Published by the city. → Hours charged against every budgeted engagement: zero against the 380 follow-up; 500 budgeted and zero charged against "Peer Review of the Internal Audit Department"; 15.25 charged against 250 budgeted for the Red Book Assessment. Records the FOAC meetings of September 11, October 9, and November 14, 2025.

Internal Audit Department Policies and Procedures Manual (as of April 20, 2026, 182 pages). Published by the city. → Page 75: the Financial Oversight and Audit Committee "request[s] external peer reviews every three years" (citing GAGAS 5.84; IIA 8.1, 8.3).

Report for the Internal Audit Current State, Maturity, and Needs Assessment — Recommendations and Road Map, Weaver and Tidwell LLP (September 2025, 70 slides). Published by the city as attachment 2 to Legistar agenda item 25-1366, City Council Work Session, November 17, 2025 — where the recorded action was "Adopted" and the result "Pass." Not posted on the Internal Audit Department's Quality Assurance page, which lists external reviews of this office back to 2007. → Maturity determination: the function "is currently at the Repeatable stage of maturity," level two of five, "with some progress into the 'Defined' level." 45 action items across 14 component areas, each with a RACI assignment across six stakeholder roles, a short-term (3–12 month) or long-term (12+ month) horizon, and an impact/effort priority. Element E: "Follow-Up and Action Plan Monitoring: Follow-up audits lack clear linkage to original audits, management accountability regarding responses and action plans is unclear." Engagement scope statement establishing the review as an assessment of structure, roles, responsibilities, and practices against IIA and GAGAS criteria — not a financial or forensic examination.

City of El Paso Internal Audit, Services & Documents page. Public. → Published audit reports through May 18, 2026, containing no follow-up to A2023-07. Quality assurance listings showing the most recent external review dated 2022.

City of El Paso press release and contemporaneous reporting, March 3, 2026. → City Council's appointment of Adrian Serrano as Chief Internal Auditor.

KFOX14/CBS4 reporting, June 19, 2025 and 2025 coverage of the assessment. Press. → The $128,000 engagement figure and the October 2024 commissioning date are drawn from press reporting; I have not examined the engagement contract. The June 19, 2025 report accurately described the preliminary results as placing the department "between levels two and three on a scale of one to five."

Calculations. The approximately 4 percent annual coverage figure is arithmetic: eight engagements against 186 scored auditable areas, both stated in the FY2025–2026 plan.

Anyone wishing to check this work can obtain every city document above from the Internal Audit Department's public page or by request under the Texas Public Information Act. Corrections are welcome and will be published.

The Benefit-Gate Gap

El Paso is negotiating a community-benefit standard for one deal that its Chapter 380 program does not require of the rest

Companion report to Memo_CBA_Binding_Language_2026-07-24.md · July 24, 2026 · Luis Ruiz, Infinitive Storytelling Same problem, different lens. The memo asks whether the City can enforce what it bought. This report asks what the City is buying at all.

The question

A community benefits agreement is, at bottom, a gate: a set of public-benefit conditions the recipient must satisfy — jobs, wages, water, workforce commitments — for the incentive to be justified. The Meta CBA is the City building that gate, carefully, for one $10 billion deal.

The question this raises is not whether the Meta gate is good. It is: does the rest of the Chapter 380 program have a gate at all?

The City's own internal audit, A2023-07 (April 17, 2024), answers it. For most of the program, the public-benefit gate is either absent or unverified. The Meta CBA is not adding a control the program has and Meta lacks. It is building, bespoke and deal-by-deal, a control the program has largely never had.

1. What a 380 is supposed to buy

The audit states the purpose plainly. The City provides incentive agreements

"for projects that will result in a net increase or retention of jobs, will add to the tax base, or will improve the economic welfare of the residents." — A2023-07

So the public consideration for a 380 incentive is three things: jobs, tax base, and resident welfare. Two of those — jobs and resident welfare — are what a community-benefits gate exists to secure. The third, tax base, follows largely from the private investment itself.

A necessary concession, stated up front. Texas law and the City's own program treat tax-base growth as a sufficient public purpose on its own — Chapter 380 authorizes incentives to "stimulate business and commercial activity," and the Texas Constitution (Art. III §52-a, which A2023-07 cites) backs it. This report does not argue that a capital-investment-only deal is illegitimate; it plainly is legal. The argument is narrower and about policy return, not statutory authority: a capex-only gate secures the weakest of the three benefits the statute names, and leaves the two that most affect residents — jobs and welfare — ungated and unmeasured.

With that concession made, the distinction is the whole report. Capital investment grows the tax base and is a valid public purpose. But jobs and resident welfare are the operational return that offsets the municipal-service strain a project creates — and they are the part that has to be gated, because unlike the building, they do not happen automatically.

2. What the program actually gates on

Here the audit is specific, and the numbers are the finding.

A. Most agreements have no jobs gate. As of August 23, 2023, the portfolio held 65 active and 111 closed 380 agreements. Of the 65 active,

"Twenty-one (21) of the 65 Incentive Agreements have a job component." — A2023-07

Forty-four of sixty-five active agreements — roughly two-thirds — carry no job-creation or -retention commitment. They gate on capital investment: the recipient builds the building. That grows the tax base, which — as conceded above — is a legally sufficient public purpose. But it secures none of the operational public benefit the statute also names — jobs, wages, resident welfare.

(These figures are the audit's August 23, 2023 snapshot. The point is not a live census of today's portfolio — templates may have changed since — but the structural baseline the City's own auditors documented: as designed and administered through the audit period, most 380 agreements gated on nothing beyond capital investment. If the program has since added benefit gates to new templates, that is the reform this report recommends, arriving early.)

B. Where a benefit gate exists, its content is soft. The one public-benefit metric in the standard 380 is the "quality jobs" threshold — jobs meeting a 380 wage-and-benefit standard. But the audit shows the definition is loose: only contractual positions must be reported, and the hours-worked standard is not fixed (the City's 380 standard is 35 hours/week, against IRS's 30 and the State's employer-defined 32–40). A gate whose unit of measure is unsettled is a gate that is hard to fail.

C. Even the benefit that is promised is not verified. This is the audit's own finding, and it is the one that matters most:

"Economic & International Development does not report to City Council the actual number of jobs created and/or retained… Only the number of jobs to be created and/or retained per each Agreement is reported." — A2023-07

The City tracks what was promised, not what was delivered. For the one public-benefit metric the program does gate on, in the minority of agreements that carry it, there is no back-end verification that the benefit ever arrived.

3. How the standard is measured — front end vs. back end

The gate metaphor becomes concrete when you ask a simple question of the statutory standard: where, in the life of an agreement, is each promised benefit actually measured? The answer explains how the program can justify incentives on "economic development" and still produce no confirmed public return.

It starts with one word: "or"

The purpose test is disjunctive:

"for projects that will result in a net increase or retention of jobs, or will add to the tax base, or will improve the economic welfare of the residents." — A2023-07

A project qualifies by satisfying any one prong. That matters because the three prongs are not equally measurable, and they are not equally earned:

The weakest prong — tax base — is the one that is both automatic and self-measuring. A deal can clear the public-purpose bar on tax base alone and never commit to the two prongs that require a real, verified public benefit.

Front end (at application and approval)

• Metrics are set here — "All Chapter 380 Agreements include a set of established metrics that the applicant must accomplish in order to receive their incentives." In practice the Compliance Division enforces two: capital investment and, where present, quality-job commitments.
• Jobs are gated only about one-third of the time — 21 of 65 active agreements carry a job component; the other 44 gate on capital investment alone.
• The job standard is soft — only contractual positions must be reported, and the hours-worked definition is unsettled (City 35/wk vs. IRS 30 vs. State 32–40).
• Financial capacity — can the applicant even deliver — was downgraded from mandatory to on-request. The 2015 policy required "solid evidence… of the business's financial stability"; the 2021 revision changed it to "Upon request, provide credible documentation." The result: 85.72% of sampled files (6 of 7) contained no financial-stability documentation (A2023-07, Findings 1–2).

So the front end asks little: qualify on one prong, prove capacity only if asked, and commit to measurable jobs only sometimes.

Back end (verification that the benefit arrived)

This is the structural hole, and it is the audit's own finding:

"Economic & International Development does not report to City Council the actual number of jobs created and/or retained… Only the number of jobs to be created and/or retained per each Agreement is reported." — A2023-07

Read prong by prong:

• Jobs — the one active public benefit that is gated is verified as promised, not delivered. The City tracks the commitment and never confirms the outcome.
• Tax base — "measured" only in the sense that the appraisal roll updates on its own; the Department does not verify it as a delivered public benefit, because it does not have to.
• Economic welfare of residents — the record shows no measurement instrument at all, at either end. It is the vaguest prong, and nothing in the audit demonstrates it is ever assessed.
• The payment and recovery controls failed as well — 60% of sampled reimbursements were paid 2–5 months late with no documented cause (Finding 4); reported incentive totals were understated ~22% by omitting waivers and TIRZ (Finding 3); and recapture, the ultimate back-end tool, was absent or structurally unable to fire ($1.35M unrecoverable, Finding 5).

The mechanism, in one sentence

The public-purpose test is satisfied on the weakest prong — measured passively and delivered automatically — while the two prongs that require real public benefit are gated inconsistently at the front end and verified essentially never at the back end.

That is how an incentive can be justified by "economic development" and still leave no measured public return on the record. The statute's "or" lets the deal qualify on tax base; the back end, where jobs and resident welfare would be confirmed, reports what was promised rather than what arrived. A gate that is checked only at entry, on the easiest prong, is not a gate — it is a turnstile.

(The back-end finding above is A2023-07's, dated April 2024. The City has since claimed remediation — addressed directly in §4, so this report is not read as ignoring it.)

4. What the subsequent reforms did — and did not — close

A2023-07 is dated April 2024. Between then and now the City has issued remediation claims and adopted new mandates, and a fair reading has to ask: is the gate gap already fixed? The honest answer is that the reforms improved tracking and audit governance, and built a front-end gate for data centers — but left the general 380 program's benefit-gate design untouched.

What the City says it fixed (Economic & International Development remediation, per A2023-07 responses):

• Back-end tracking — the strongest claim. "As of January 1, 2025, the Department transitioned the Incentive Agreement Portfolio to Monday.com as the centralized database for tracking actual job counts, fee waivers, TIRZ, and rebate activity," shared with Council "at least annually." If delivered as described, this directly addresses the promised-vs-actual back-end finding in §3. It should be conceded, not ignored. What it does not do is create a benefit obligation where none was set — it tracks whatever gates exist more accurately.
• Front-end financial vetting — partial. ED reported that "as of January 1, 2025, [it] has 22% documented" of applications pre-vetted for financial capacity against SEC/IRS/Secretary of State/Comptroller/BLS databases. 22% is progress on Findings 1–2, not closure.
• A revised 380 Incentive Policy with a risk-assessment framework was promised "before June 2026" — treat as promised, not confirmed enacted (same posture as the general 380/381 rewrite).

The 2026 mandates operate one level up — on the audit function, not the gate:

• The April 2026 Internal Audit Policies and Procedures Manual governs how audits are selected and conducted (risk weighting, IIA/GAGAS conformance). It names "380 Agreements" as an audit area, but it does not set 380 gate terms.
• The June 2026 Internal Audit Charter expansion strengthens audit independence and risk frameworks. Also audit-function, not gate design.

The one genuine front-end gate the City built — and its scope:

• In July 2026 the City adopted a data-center-specific policy framework (stricter oversight, no future incentives for hyperscale data centers). This is a real front-end gate. It proves the City can design one when it chooses to. It is confined to a single sector.

What none of the above touches — the persisting gap:

• Most agreements still gate on capital investment alone (44 of 65 in the audit population); better tracking does not add a benefit obligation to a deal that never carried one.
• The disjunctive "or" is unchanged — a project can still clear the public-purpose test on tax base alone.
• There is still no program-wide requirement of a measurable public benefit on ordinary 380s.

The synthesis, updated for the reforms: the City has improved how it counts (Monday.com), how it audits (2026 manual and charter), and how it gates one sector (the data-center framework). It has not changed how it gates the general program. Better back-end tracking of a front-end gate that mostly does not exist still leaves most 380s without a measurable public-benefit requirement. And because the City demonstrably can build such a gate — it just did, for data centers — the general-program omission now reads as a choice, not a capacity limit. That is the narrower, current, and harder-to-rebut version of this report's thesis.

5. The gap, stated

Put the three together and the shape is clear:

The City is assembling, for a single data center, the community-benefit gate the audit shows the general program lacks. That is not wrong — a $10B / 1GW project earns bespoke scrutiny. But it means the CBA is a patch on a program-wide absence, and the moment to notice that is now, while the gate is being drawn for the first time and could be drawn as a program standard rather than a one-off.

This is the same finding the companion memo reaches from the enforcement side. There, a recapture clause existed and did not fire because the components that got built satisfied it. Here, a benefit gate mostly does not exist, and where it does, delivery is not measured. Both are the same underlying condition: the public's half of the bargain is written loosely and checked rarely.

6. The recommendation

It is the same baseline the companion memo proposes in its Section 4, viewed from the benefit side. Two additions belong in any program-wide 380 standard:

1. A measurable public-benefit gate on every agreement, scaled to incentive size — a jobs, wage, or defined public-improvement commitment, not capital investment alone. Capex grows the tax base; it is not a substitute for the jobs-and-welfare benefit the statute names.
2. Back-end verification of delivered benefit, reported to Council — the actual jobs created and retained, measured against the promise, for every agreement carrying a benefit commitment. This closes the exact gap A2023-07 documents: promised is tracked; delivered is not.

Scale dictates structure — the gate scales down, it is not copy-pasted. A hyperscale data center's gate is grid, water, and workforce; a retail or mixed-use deal's gate is phased completion and a wage floor; a logistics deal's is job density. What every tier shares is a measurable public benefit, verified after the fact — that is the floor, and it is what the audit shows is missing across most of the program. Demanding the Meta apparatus of a strip-mall infill would suffocate small local development; that is not the recommendation, and the companion memo's Section 4 sets out the same baseline-plus-sector-modules structure so the concept scales without the burden.

7. Scope and precision

• This report is a program-level reading of A2023-07. The audit reports the aggregate (21 of 65 active agreements carry a job component); it does not support attributing gate status to any individual agreement or recipient, and none is asserted here.
• "Community-benefit gate" is used in the plain sense — enforceable public-benefit conditions on the incentive — not as a reference to any single named instrument.
• No intent is imputed to any party. The absence of program-wide benefit gates is a policy-design fact, documented by the City's own auditors, not a claim about anyone's motive.
• Register throughout: auditor, not prosecutor. The City established the Compliance Division in 2012 and commissioned A2023-07 itself; this report reads the City's own record back to it.

Sources

A2023-07 was issued by the City of El Paso Internal Audit Department, April 17, 2024. This report and the companion memo are governance-audit analysis, not legal advice. Remediation claims are the City's own statements as recorded in A2023-07; their delivery is not independently verified here.

Audit Report: "A bank email after January 6 opens a much older question" (ToreSays — Operation CHAOS / EO 12333 / J6 bank thresholds)

Date: 2026-09-26 Format: narrative_article Assessment Mode: autonomous (external content) Instrument: audit-claims@5.0 Total Claims: 40 (after deduplication)

Bottom line

The body of this piece is well sourced. All 30 fully sourced claims hold up against the primary documents: the House Judiciary reports, Rockefeller Commission ch. 11, Church Committee Book III, the amended EO 12333 and the 2008 AG Guidelines. The problems are elsewhere:

1. What the frame says vs. what the body says. The body states plainly that the record does not connect the CIA to the bank search. The opening sentence, a section heading, the closer, and the next-installment teaser ("The Paper Path From Langley to Your Bank Statement", "the formula… needed only a new emergency") all assert that connection anyway.
2. Omissions that cut the other way. Several facts from the very sources it cites would weaken its argument, and the article leaves them out. The sharpest one: the FBI's own Domestic Terrorism section chief had the BoA data pulled from FBI systems because it "lacked allegations of federal criminal conduct." That answers the article's own question: "whether people who are cleared leave the file."
3. Embellished details in the CHAOS section. "Drawerful" of letters, the Akron comparison, "seven years", "the people closest… flinched."

Pre-Screen (Tavistock — Pass 1)

• NLP moves: 53 (18.3 / 1k words). Tavistock moves: 19 (6.5 / 1k). Sycophancy markers: 0/6. Meta flag: false. Truncated: no (2,903 words).
• Mostly false positives. 13 of the 14 regression_anxiety hits are the proper noun "CHAOS". The 3 dependency hits are "General" (Attorney General / general counsel). The 36 nominalizations are mostly "the Commission"-type noun phrases. Instrument note: the detector should exempt proper nouns / operation names.
• The detector missed the real pattern (frame vs. body, accusation-in-a-mirror) because that pattern is structural, not lexical. The audit's findings govern.
• Raw: matrix · excerpts · json

Document Notes

• Deduplication: no repeated paragraphs. Footer (tip jar, series list) stripped.
• Missing content: three images/charts are referenced ("The loop…", "The CHAOS record, 1967–1975…", "All bars to one scale…") but were not supplied. Their figures can't be audited.
• Child protection: none.
• Meta-manipulation: flagged as tactic-present (see Document-Level Analysis). No intent inferred.
• Named individuals: historical officials (Ober, Karamessines, Angleton, Helms), Judge Cobb, and Tom Fitton, all accurately described. Peter Sullivan, the FBI sender, is not named in the article.

Summary Statistics

• Sourcing Rate: 75% (30 sourced / 40 total). Primary metric for this format. 6 of the 10 unsourced are partials, flagged for human keep/throw.
• Accuracy Rate: 100% (30 / 30 sourced claims verified against primary documents)
• Sycophancy / Authority Deference: 1
• Emotional Manipulation: 7
• Logical Fallacies: 4
• NLP / Linguistic Persuasion: 6
• Fifth-Gen Warfare Patterns: 3 (document level)
• Legal Flags: 3 (all low)
• High-Risk Flags: 2 (editorial/accuracy risk, not legal)

Claims Audit

Bank of America / FBI (claims 1–6, 37)

Operation CHAOS (claims 7–28)

Legal framework (claims 29–32, 35)

CIA / January 6 (claims 33–34, 36, 38)

Frame and teaser (claims 39–40): high risk

Claim 39. Teaser: "a foreign hook supplied within ten days… four streams of data… an intelligence center helping write what the banks were told to look for… cameras already rolling before the crowd moved."
Sourcing: ✗ none. Tactics: presupposition, emotional framing, information fragmentation. Legal: low (no named party). High risk: editorial. Four factual assertions stated as established, with the evidence deferred to a later installment. Rec: Seek source.

Claim 40. Title and closer: "The Paper Path From Langley to Your Bank Statement"; "this playbook"; "The formula did not need to be reinvented. It needed only a new emergency."
Sourcing: ✗. Internal inconsistency: directly contradicts the article's own conclusion ("The public record available today does not join those last two events into one CIA-directed operation") and its caveat ("The existence of a route is not proof someone traveled it"). Tactics: presupposition, embedding, false analogy, pacing-and-leading. Rec: Rephrase.

High-Risk Flags

• #39, #40. The frame asserts what the body disclaims. This is a credibility and accuracy risk, not a defamation risk: no private individual is named, and government entities are not libel plaintiffs.

Unsourced Claims

6, 16, 18, 19, 20, 23, 27, 35, 39, 40. The partials (6, 16, 20, 23, 27, 35) are flagged for human keep/throw.

Per-Claim Tactic Summary

• Emotional framing (7): #14 "the part that should stop you," #16 "after the program was dead," #18 superlative, #20 "drawerful," #23 Akron "every man, woman, and child," #27 "flinched," #39 "cameras already rolling."
• Logical fallacies (4): #6 false precision (sequence), #16 false precision (years / count of "no"s), #23 false equivalence (names incl. organizations → people), #40 false analogy (CHAOS "formula" → J6).
• Authority / credential (1): "My FOIA requests pursue the adjoining paper trail." No request, agency or date is shown.

Document-Level Tactic Analysis

• Tribal signaling: the sign-off "It's not the story they tell you that is important. It's what they omit." sets up an unnamed they against the reader.
• Accusation in a mirror (tactic-present, not intent): the piece accuses "they" of shaping stories by omission, and it teaches that a label "ends the argument without answering it." Meanwhile it:
  • omits that the FBI pulled the BoA data (claim 4);
  • omits Hill's "no directive from the FBI" (claim 4);
  • omits FinCEN's convening role (claim 2);
  • omits "None was purely domestic" (claim 21);
  • omits the lives-endangered clause of §2.6(c) (claim 35);
  • closes on the labels "playbook" and "formula."
• Information fragmentation: the most consequential claims are deferred to the next installment, where this piece gives no way to check them.
• What the article does well: unusually explicit hedging. It says the SAR is "not an accusation, much less proof," that the FBI "does not need a CIA request," and that "an acknowledgment, a delay, or a withholding is not proof." These hedges are real and accurate. The problem is that the frame doesn't obey them.

Connective Tissue NLP Summary

• Opening sentence (highest leverage): "A bank email after January 6 opens a much older question: how can a foreign intelligence inquiry become a file on Americans?" To read on, the reader has to accept that the 2021 email involves a foreign-intelligence inquiry. The cited record shows an FBI Counterterrorism Division / FinCEN / bank process with no foreign-intelligence requirement. The article later concedes this. The link is planted in line one and disclaimed in the middle.
• Transitions: "We have seen that happen before" asserts the two events are the same kind before comparing them. "There is another reason to read the order carefully" places the bomb techs next to the bank search, which implies a connection. "Where does the road lead?" moves from the disclaimer straight into the Langley teaser.
• Qualifiers: "the missing classification" presupposes a withheld record. "even one you find too broad" presumes the reader's view. "Readers deserve that precision" is self-credentialing placed just before the imprecise teaser.
• Sequencing: careful body, assertive frame. Each hedge builds credit that the closer spends.
• Term drift: "names" → "every man, woman, and child"; "foreign intelligence inquiry" (CHAOS) → applied implicitly to the 2021 bank email; "CIA personnel were described" → "the CIA's January 6 deployment."

Recommendation Summary

• Publish as-is: 27 claims (the documentary spine is solid). Optional tightening: add FinCEN context to #2, "two Presidents" in #14, and drop "sometimes" in #25.
• Publish with context: 4 (#4, #19, #34, #38).
• Rephrase: 8 (#6, #16, #18, #20, #23, #27, #35, #40).
• Seek source: 1 (#39).
• Single highest-value fix: add the Jensen "pull" fact next to claim 4. It is in the same March 2024 report the article already cites, and it directly answers the question the piece says matters most.

Sources verified: House Judiciary Dec 2024 report · House Judiciary Mar 2024 interim report · Rockefeller Commission ch. 11 · Church Committee Book III, CHAOS · EO 13470 amendments to EO 12333 · AG Guidelines 2008 · Judicial Watch ATF records · Judicial Watch Sept 24, 2026 · The Blaze · Akron population history

Audit Report: Santa Teresa Strategic Plan — Strategic Assessment Report

Date: 2026-06-21 Format: quantitative_advocacy (first-class as of audit-claims v5.1 + Verity QIL v1) Assessment Mode: autonomous Instrument: audit-claims@5.1 + Verity QIL v1 Source: STSP_Strategic Assessment Report, Wilson & Company / NMSU Arrowhead Center, dated March 2025 (file 06-02-25) Lead Metric: Quantitative Integrity Score (not sourcing rate — this is a numbers-first funding case)

Orientation. This audit is pointed at the document on the author's behalf. Every finding is framed claim → vulnerability → how a hostile reviewer attacks it → how to harden it. The goal is to make the ~$1.26B Phase 1 funding case survive adversarial review, not to tear it down. Where a number is sound, it is marked defensible and left alone.

Quantitative Integrity Summary

• Quantitative Integrity Score: 56% (5 of 9 material numeric claims pass all applicable gates)
• Material numeric claims: 9 (scored) + 1 document-integrity inconsistency (N9)
• Keystone assumptions: 3 (listed below)
• High-vulnerability numeric claims: 4 (N1, N2, N4, N5)
• Numeric Armor findings: 6
• Internal inconsistencies (numeric): 3 (CAGR window; table numbering; Phase-1 cost cross-foot)

Reported separately, never averaged into the QIS: the 3 keystone fragilities and 6 Numeric Armor findings. A single load-bearing keystone with a reconciliation defect (N2) matters more than the aggregate score, and burying it in a mean would defeat the purpose of the layer.

Keystone Assumptions

The funding case rests on three inputs. If any one moves, the headline numbers move with it.

K1 — 18.5% CAGR extrapolated 2024–2028. (p.111–112, "Continued Growth vs. Halted Development") The entire job- and output-gap case is built by projecting recent industrial-employment growth forward five years at 18.5% compounded.

• Figure / location: "compound annual growth rate of 18.5%," used as the Continued-Growth scenario driver (p.111, p.112).
• ±20% sensitivity: the headline 9,755-job and $2.44B-output gaps scale almost linearly with this rate. Cut it to ~14.8% and you roughly halve the stakes; the appropriation ask loses its proportion.
• Provenance: model input (asserted historical rate carried forward).
• Reconciliation defect: the same 18.5% figure is attributed to "2019–2023" (p.111) and "Between 2020 and 2023" (p.112). A different base window changes the CAGR. (Verified against the source text — both windows appear, both labeled 18.5%.)
• Vulnerability: high. → reconcile + disclose_assumption + add_scenario.

K2 — Halted-Development floor. (p.112, Table 15) The headline difference is the gap between the 18.5% trend (ceiling) and a "Halted Development" scenario (floor) in which growth stops because infrastructure isn't funded. No mid-case sits between them.

• Figure / location: Halted Development 2028 = 7,179 total jobs / $2.17B output vs. Continued Growth = 16,934 jobs / $4.61B output (Table 15).
• ±20% sensitivity: the entire 9,755-job / $2.44B "difference" is the floor-to-ceiling spread. A less pessimistic floor (e.g. growth slows rather than halts) shrinks the difference directly.
• Provenance: model construction; the floor's growth assumption is not quantitatively justified.
• Vulnerability: high. → add_scenario.

K3 — 80% of economic benefit accrues to New Mexico. (p.111) Determines how much of the modeled impact NM can claim for a state appropriation.

• Figure / location: "roughly 80% of the economic benefits… accrues in New Mexico," carried from "NMSU's previous studies."
• ±20% sensitivity: a 64% allocation would cut NM's claimable share by a fifth across every downstream figure.
• Provenance: self-referential (prior NMSU studies, not provided in this document).
• Vulnerability: medium. → cite_source.

Numeric Claims Audit

N-Claim #1 — Continued-Growth headline

Text: "if the required infrastructure investments are in place… Santa Teresa's economic output would grow to $4.6 billion by 2028, adding more than 10,300 direct jobs and nearly 17,000 total jobs" (Exec. Summary, p.ES-2)
Figure: $4.6B output; 10,300 direct / ~17,000 total jobs by 2028
Decomposition: input (18.5% CAGR + job-to-sqft ratio on 6,123,234.40 sq ft) → method (IMPLAN + MRIO extrapolation) → output ($4.61B / 16,934 jobs)

• D1 Provenance: model-output-as-fact — stated in declarative voice in the executive summary; method named later (p.111) but not at the point of assertion.
• D2 Keystone: yes — depends on K1.
• D3 Baseline: trend-extrapolation (the ceiling scenario).
• D4 Attribution: —
• D5 Metric selection: flag — leads with gross economic output ($4.6B); the more conservative value-added figure ($1.75B, Table 15) is not led with.
• D6 Precision: rounded in prose ($4.6B); fine here.
• D7 Reconciliation: consistent with Table 15.
• Vulnerability: high
• Recommendation: requalify_metric (+ disclose_assumption)

N-Claim #2 — 18.5% CAGR driver

Text: "The projected growth rate of employment in Santa Teresa for 2024–2028 is based on the historical compound annual growth rate of 18.5% observed from 2019–2023." (p.111) / "Between 2020 and 2023, industrial employment… grew at a compound annual growth rate (CAGR) of 18.5%." (p.112)
Figure: 18.5% CAGR
Decomposition: input (employment counts at two endpoints) → method (CAGR) → output (18.5%, then extrapolated 5 yrs)

• D1 Provenance: asserted / model input.
• D2 Keystone: yes — K1.
• D3 Baseline: trend-extrapolation; cherry-picked-window risk — the base years are stated inconsistently, and a 4-yr vs 3-yr window changes the rate.
• D6 Precision: a single-decimal rate presented as a hard driver.
• D7 Reconciliation: flag — "2019–2023" vs "2020–2023" for the same figure.
• Vulnerability: high
• Recommendation: reconcile + disclose_assumption + add_scenario

N-Claim #3 — EROI 196% / 55%

Text: "The EROI for the infrastructure improvements… is estimated at 196% using economic output estimates, and 55% using value added production estimates." (p.113)
Figure: EROI 196% (output) vs 55% (value-added); 3.56× spread
Decomposition: input (net benefit stream vs Phase-1 cost stream) → method (IRR on net economic benefit) → output (196% / 55%)

• D1 Provenance: model-output; both metrics disclosed (the honest move).
• D5 Metric selection: flag — the 3.56× spread is led by the flattering 196%; 55% is the figure tied to value actually added.
• D7 Reconciliation: internally consistent with the EROI input tables.
• Vulnerability: medium
• Recommendation: requalify_metric (lead 55%, or present both with equal weight)
• Gate status: passes QIS gates — discloses both figures, reconciled, keystone (18.5%) named in-section. The flag is presentational, not a gate failure.

N-Claim #4 — 23,729 jobs lost since 2021

Text: "since 2021, the statewide economic impact of major projects choosing to locate elsewhere is estimated to include the loss of 23,729 jobs (15,922 direct jobs), $8.5 billion in economic output… Recently, eight major projects chose against Santa Teresa's location." (Exec. Summary, p.ES-2)
Figure: 23,729 jobs / $8.5B output "lost"
Decomposition: input (8 projects × avg 1,990 jobs / $2.7B capital) → method (impact model on foregone projects) → output (23,729 jobs / $8.5B)

• D1 Provenance: self-referential / asserted — "NMSU and MVEDA have analyzed"; no external dataset or per-project method shown.
• D4 Attribution: flag — the loss is pinned on Santa Teresa's infrastructure deficiency. Competing explanations (Texas incentives, tax structure, workforce, site readiness) are not ruled out. Post-hoc/false-cause in quantitative clothing.
• D6 Precision: "23,729" and "15,922" stated to the single job off a modeled foregone-project count.
• Vulnerability: high
• Recommendation: reframe_causal (present as correlation, or add project-by-project reason coding)

N-Claim #5 — 9,755-job / $2.44B difference

Text: Table 15 — Total Jobs: Halted 7,179 vs Continued 16,934, difference −9,755; Economic Output: Halted $2,171,016,719 vs Continued $4,614,303,648, difference −$2,443,286,929. (p.112)
Figure: 9,755 jobs; $2,443,286,929
Decomposition: input (two scenario outputs) → method (subtraction) → output (the "difference" headline)

• D2 Keystone: depends on K1 + K2.
• D3 Baseline: manufactured difference — floor (Halted) vs trend (Continued), no mid-case. The headline is the maximal spread.
• D6 Precision: $2,443,286,929 — to the dollar (see N6).
• D7 Reconciliation: consistent across Table 15 and the EROI tables.
• Vulnerability: high
• Recommendation: add_scenario (a realistic mid-case band makes the hardened number more persuasive than the maximal one)
• Gate status: passes QIS gates (model method disclosed, reconciled, assumption named) yet remains high-vulnerability via D3 — the layer working as designed: a gate-clean claim that a reviewer still attacks.

N-Claim #6 — to-the-dollar / to-the-hundredth precision

Text: "$2,443,286,929" (Table 15); "6,123,234.40 square feet of existing industrial space" (p.111)
Figure: $2,443,286,929; 6,123,234.40 sq ft

• D1 Provenance: the dollar figure is model-output; the sq-ft figure is Wilson & Company's own land-use measurement.
• D6 Precision: false precision — a modeled estimate carried to the dollar; floor area to the hundredth of a square foot. The significant figures imply a confidence the inputs (an extrapolated growth rate) cannot support.
• D7 Reconciliation: see N10 — the Phase-1 cost table built from figures at this precision does not cross-foot.
• Vulnerability: low
• Recommendation: restate_precision (round to honest significant figures: ~$2.44B; ~6.12M sq ft)

N-Claim #7 — $39.5B international trade "historic high"

Text: "International trade through the Santa Teresa Port of Entry in 2024 reached $39.5 billion, a historic high. Imports from Mexico… 59.4% and equal to $23.5 billion; exports to Mexico reached $16 billion." (p.114)
Figure: $39.5B (imports $23.5B + exports $16B)

• D1 Provenance: sourced — verifiable POE trade data.
• D5/D6: clean.
• D7 Reconciliation: $23.5B + $16B = $39.5B — reconciles exactly.
• Vulnerability: none
• Recommendation: defensible — the layer working in the other direction. The number is sound; it carries only light emotive armor ("historic high"), noted but not penalized.

N-Claim #8 — 80% benefit-to-NM allocation

Text: "roughly 80% of the economic benefits of activity in the Santa Teresa Industrial Parks accrues in New Mexico." (p.111)
Figure: 80% allocation to NM

• D1 Provenance: self-referential (prior NMSU studies, not provided).
• D2 Keystone: yes — K3; governs NM's claimable share for the appropriation.
• Vulnerability: medium
• Recommendation: cite_source (name and date the prior NMSU studies; ideally show the allocation method)

N-Claim #9 — table-numbering instability (document-integrity inconsistency)

Text: The List of Tables jumps Table 15 → Table 17 → Table 18 (no Table 16 listed). In the body, "Table 16" labels both the Infrastructure Investment table (p.112) and the EROI Economic-Output table (p.113); the same EROI tables are then cross-referenced in prose as "Table 17 and Table 18."
Figure: table labels 16/17/18

• D7 Reconciliation: flag — the same label ("Table 16") is used for two different tables, and the EROI tables carry different numbers in the body vs. the ToC vs. the cross-reference sentence. Not a numeric error, but a citation-integrity defect a careful reviewer will notice and use to question the document's rigor.
• Vulnerability: low (not scored in QIS denominator; counted as an internal inconsistency)
• Recommendation: reconcile (renumber tables consistently across ToC, labels, and cross-references)

N-Claim #10 — Phase-1 cost $1,260,086,404 (the ask)

Text: Table 16 — Total Phase-1 infrastructure: Water $326,530,555 + Wastewater $624,448,055 + Transportation $274,500,000 + Drainage $34,607,794 = $1,260,086,404. (p.112)
Figure: $1,260,086,404
Decomposition: input (engineering cost estimates by category/year) → method (sum) → output ($1.26B)

• D1 Provenance: asserted — Wilson & Company engineering cost estimates (the firm's own).
• D6 Precision: false precision — planning-grade cost estimates carried to the dollar.
• D7 Reconciliation: flag — the table does not cross-foot. Summing the four annual column-totals gives $1,260,086,407; summing the four category row-totals gives the stated $1,260,086,404 — a $3 discrepancy traced to rounding in the Wastewater (−$1) and Drainage (−$2) subtotals. Immaterial in dollars, but it undercuts the to-the-dollar precision posture. (Verified by recomputation.)
• Vulnerability: low_medium (it is the appropriation denominator)
• Recommendation: restate_precision + reconcile (present as "~$1.26B," make subtotals cross-foot)

Numeric Armor Cross-Reference

Where a soft number is carried by hard language — the prose doing persuasive work the number cannot.

1. "NMSU estimates that if the required infrastructure investments are in place… output would grow to $4.6 billion" + $4.6B projection → conditional-sold-as-forecast. The hedge ("if… would") is present but seated in declarative, authoritative voice so it reads as a prediction.
2. "Economic impact analysis… underscores the urgency of infrastructure investment" + the modeled 2024–2028 output → urgency-on-projection. Evaluative urgency layered onto an unrealized model output.
3. "These statistics underscore the critical need… to ensure continued growth" + the 23,729-jobs-lost figure → urgency-on-projection (urgency attached to a self-referential loss estimate).
4. "robust employment growth" / "more robust infrastructure" + the 18.5% extrapolation → emotive-on-estimate. Affective framing pre-empts scrutiny of the growth assumption.
5. "$39.5 billion, a historic high" + the (sound) trade figure → emotive-on-estimate. Here the number is verifiable; the armor is light and the claim still earns defensible. Noted to show the pattern is detected regardless of whether the underlying number holds.
6. IMPLAN/MRIO modeling + to-the-dollar figures ($2,443,286,929; 6,123,234.40 sq ft) presented as rigorous, study-based → precision-as-authority. The precision itself is the credibility signal. Cross-references the aggregate-as-headline pattern: gross output ($4.6B) leads in prose while value-added ($1.75B) sits unhighlighted in Table 15.

Hostile-Reviewer Brief

Severity-ranked. This is the section to hand the author.

1. Keystone — 18.5% CAGR (N2 / K1). Attack: "Your entire loss-and-gain case assumes five straight years of uninterrupted 18.5% growth, and your own report can't agree on the base years — is it 2019–2023 or 2020–2023? Pick one, because the rate moves with it." Fix: Reconcile the window, disclose the extrapolation as an assumption at the point of assertion, and add a conservative mid-case so the ask survives the first hostile question.

2. Attribution — 23,729 jobs lost since 2021 (N4). Attack: "You attribute 23,729 lost jobs to Santa Teresa's infrastructure, but Texas out-incentivized you on most of those eight projects. Prove infrastructure — not incentives, tax, or workforce — was the deciding factor." Fix: Present the figure as correlation, or add project-by-project reason coding showing infrastructure was decisive.

3. Scenario design — the 9,755-job / $2.44B gap (N5 / K2). Attack: "This is the gap between a worst case and a best case you built yourselves. Where's the realistic middle? Without it, the headline is a maximum, not an estimate." Fix: Insert a mid-case band between Halted and Continued. The hardened number is more persuasive than the maximal one.

4. Metric selection — $4.6B / 196% led, $1.75B / 55% buried (N1 / N3). Attack: "You lead with $4.6B gross output and a 196% EROI. The value-added figures — $1.75B and 55% — are 2.6× and 3.6× smaller and are what the state actually captures. Why aren't those the headline?" Fix: Lead with value-added alongside output, or requalify the headline so the conservative figure carries equal weight.

5. Precision / integrity (N6, N9, N10) — low severity, high credibility cost. Attack: "You present $2,443,286,929 and 6,123,234.40 square feet as if measured, your Phase-1 cost table doesn't cross-foot by $3, and your tables are misnumbered. If the easy arithmetic is loose, why should I trust the model?" Fix: Round modeled figures to honest significant figures, make the cost table cross-foot, renumber tables consistently.

Calibration & Method Notes

• This run reproduces the QIL v1 Appendix A calibration on the same document and independently verified each figure against the source text (page footer labels ES-2, 111, 112, 113, 114).
• Two catches beyond the calibration appendix: (a) the table-numbering instability (N9), and (b) the $3 cross-foot mismatch in the Phase-1 cost table (N10), both confirmed by recomputation.
• QIS = 5/9 = 56%. Gates per claim: provenance-clean, reconciled, keystone-disclosed. Passes: N3, N5, N6, N7, N10. Fails: N1 (model-output-as-fact, keystone not disclosed at assertion), N2 (reconciliation defect), N4 (self-referential provenance, method not shown), N8 (self-referential provenance). N5 illustrates the design point: gate-clean yet high-vulnerability.
• Schema status: quantitative_advocacy is now first-class — added to audit-claims.schema.json (format enum), the SKILL Step-0 format table, the v5 contract enum, the JSON template, and Step 3 (lead-metric routing) under the v5.1 additive revision. The companion JSON sets manuscript.format: "quantitative_advocacy" directly and validates against the updated schema. This is an additive enum widening plus optional QIL output blocks; it changes no per-claim weight or rule, so existing 5_Data_v2/*.audit.json stay conformant and need no re-run (they gain QIL blocks only when naturally re-audited).

Generated by Verity QIL v1 (bolt-on to audit-claims@5.0). Numbers verified against STSP_Strategic Assessment Report. Findings harden the case; they do not detonate it.

Audit Report: Meta reps meet with El Paso data center critics, city leaders as proposal asks more investments of tech giant

Source: El Paso Matters (Diego Mendoza-Moyers), June 18, 2026 — URL not included in the source file provided to this audit. Date Audited: 2026-08-13 Format: narrative_article Assessment Mode: autonomous Skill Version: audit-claims@5 (canonical; see Document Notes on skill drift) Total Claims: 26 (after deduplication)

Document Notes

Format: narrative_article — sourcing rate is the primary metric.

Deduplication: No repeated content. Full article text as provided; 1,795 words. No dedup required.

Completeness: No genuine "missing content" gaps (no chart, table, or example referenced but absent from the document). The "READ MORE" cross-link and the "Who benefits from the AI boom?" call-out box are self-promotional/navigational inserts, not claims.

Child protection: No minors named or discussed. No exposure.

Skill version drift: The audit-claims skill synced into this remote session is v3 (older tactic taxonomy, no pre_screen block). This audit was instead written to conform to the project's canonical v5 schema and validator (audit-claims.schema.json, validate_audit_output.py), matched to the shape of the most recent conformant audit already in 0_Inbox (KVIA Borderplex Connect). The JSON companion validates clean against the schema. Recommend closing the v3/v5 drift noted in your feedback_audit_claims_skill_v1.md memory so future remote-session audits pick up v5 automatically.

Pre-screen (Tavistock Pass 1): Run locally via the device bridge against the article text. 26 NLP lexicon hits, 5 Tavistock moves, word count 1,795 (not truncated). Density: 14.5 NLP/1k words, 2.8 Tavistock/1k words. No sycophancy markers, no conditioning hits, no meta-manipulation flag. Most hits were mechanical repetition (Meta's own term "restoration" used 4x as a term-of-art, generic nominalizations like "the deal"/"the proposal") rather than manipulative framing on inspection. Step 2B below independently confirms 4 of the 26 as genuine connective-tissue moves. Pre-screen counts are diagnostic only and are not summed into the stats below.

Notable finding: This is one of the more balanced documents in your corpus by format — it quotes 8+ named parties spanning company, city, utility, union, and resident-critic positions in comparable proportion, rather than a single-source frame. It also contains a rare self-correcting statistic: Balliew's "<1%" water-demand claim (Claim 3) is independently checkable against two other figures cited in the same article and holds up on the math.

Summary Statistics

• Sourcing Rate: 61.5% (16 sourced / 26 total) — primary metric for this format
• Partial (counts as unsourced): 5 claims (#7, #10, #11, #13, #16)
• Accuracy Rate: 100% of sourced, checkable claims (no misreadings found); Claim 3 cross-validated against Claims 4–5 within the document
• Sycophancy Detected: 1 instance
• Emotional Manipulation: 1 instance
• Logical Fallacies: 4 instances
• NLP / Linguistic Persuasion: 4 instances
• Fifth-Gen Warfare Patterns: 0 (document level)
• Legal Flags: 4 (all rated low)
• High-Risk Flags: 0

Claims Audit

Claim #1

Text: "Other participants included El Paso Electric CEO Kelly Tomblin, El Paso Water CEO John Balliew, District 4 city Rep. Cynthia Boyar Trejo, Deputy City Manager Robert Cortinas, Meta executives Ana Martinez and Darija Cosic, City Attorney Karla Nieman, Northeast resident John Justice, attorney Veronica Carbajal, union leader Eddie Trevizo, El Paso Chamber CEO Ricardo Mora, Greater El Paso Association of Realtors governmental affairs director Ernesto Garcia, and a representative from District 3 city Rep. Deanna Maldonado-Rocha's office."
Sourcing: ✓ sourced — reporter's direct observation
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #2

Text: "Boyar Trejo and Johnson said they will hold more community discussions with Meta."
Sourcing: ✓ sourced — reporter's direct account
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #3

Text: "Balliew of the water utility said Meta's water consumption will increase overall water demand by less than 1%. 'The Meta water consumption, in the grand scheme of things, is not as large as it is being portrayed,' Balliew said."
Sourcing: ✓ sourced — on-record statement, El Paso Water CEO John Balliew
Tactics: NLP — presupposition ("as it is being portrayed" implies an unnamed exaggerated portrayal without naming who is doing the portraying)
Accuracy Note: The <1% figure is independently checkable against Claims 4–5 in this article (500,000/108,000,000 ≈ 0.46%; 500,000/155,000,000 ≈ 0.32%) and holds up.
Legal Exposure: none
Recommendation: Publish with context

Claim #4

Text: "The utility last fiscal year supplied about 108 million gallons of water to customers per day – and supplied 155 million on the peak demand day last June."
Sourcing: ✓ sourced — Balliew, utility operational data
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #5

Text: "Meta's facility will use somewhere between 400,000 to 500,000 gallons per day on average."
Sourcing: ✓ sourced — Balliew, presumably from Meta's water service application
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #6

Text: "To plan conservatively, El Paso Water assumes the same conditions from the most intense drought period on record – El Paso received a record-low amount of river water in 2013 – will exist every year going forward."
Sourcing: ✓ sourced — Balliew, utility planning methodology; 2013 drought is public hydrological record
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #7

Text: "'If we were building houses out there, a normal subdivision, we would be using about the same amount of water. Nobody would be talking about it,' Balliew said."
Sourcing: ~ partially sourced (counts as unsourced) — speaker is named and accurately quoted, but no comparative water-use figure for an equivalent subdivision is cited
Missing Content: No water-use figure for a comparable subdivision
Tactics: Logical Fallacy — false equivalence (data center vs. subdivision water use, eliding differences in community benefit and permanence)
Legal Exposure: none
Recommendation: Rephrase

Claim #8

Text: "...referring to the $300 million water treatment plant under construction in the Lower Valley. The plant will convert wastewater into 10 million gallons of potable water daily."
Sourcing: ✓ sourced — Balliew, referencing the Pure Water Center (public infrastructure project)
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #9

Text: "Meta has said it will replenish twice as much water as its facility uses."
Sourcing: ✓ sourced — attributed directly to Meta, properly hedged
Tactics: None (flagged for context: this is Meta's own forward pledge, not an independently audited figure)
Legal Exposure: none
Recommendation: Publish with context

Claim #10

Text: "sensor-based irrigation technology Meta says will allow farmers to use water more efficiently and save 105 million gallons of water annually"
Sourcing: ~ partially sourced (counts as unsourced) — attributed to Meta, methodology not shown
Missing Content: Underlying study/dataset for the 105M-gallon figure
Tactics: Logical Fallacy — false precision (exact figure implies rigor with no shown methodology)
Legal Exposure: none
Recommendation: Seek source

Claim #11

Text: "drip irrigation to reduce usage by 71 million gallons annually"
Sourcing: ~ partially sourced (counts as unsourced) — same issue as Claim 10
Missing Content: Underlying study/dataset for the 71M-gallon figure
Tactics: Logical Fallacy — false precision
Legal Exposure: none
Recommendation: Seek source

Claim #12

Text: "Meta will also work with the nonprofit DigDeep to provide first-time running water to 28 homes in El Paso County"
Sourcing: ✓ sourced — attributed to Meta; DigDeep is a real, identifiable nonprofit and the figure is specific and checkable
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #13

Text: "She added that an independent third-party firm would verify Meta's water restoration project data. 'Within our sustainability report, we will also report on water consumption for each of our operational data centers. And we'll also have a volumetric water benefit accounting report...'"
Sourcing: ~ partially sourced (counts as unsourced) — attributed to Cosic; no verification firm named
Missing Content: Name of the third-party verification firm
Tactics: NLP — nominalization ("an independent third-party firm" is an unnamed, uncheckable actor)
Legal Exposure: none
Recommendation: Seek source

Claim #14

Text: "The data center will require 1 gigawatt of electricity, much of which will be supplied by natural gas generators that will emit pollutants as well as greenhouse gas emissions."
Sourcing: ✗ unsourced — no source named in this article
Tactics: None (plausibly drawn from prior El Paso Matters coverage or public filings, but this document doesn't support it standalone)
Legal Exposure: low — misleading-statistic risk (unsupported magnitude claim about environmental impact)
Recommendation: Seek source

Claim #15

Text: "El Paso Electric didn't immediately respond to questions about the climate emissions the utility expects the McCloud facility to release."
Sourcing: ✓ sourced — reporter's own account of the reporting process
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #16

Text: "The power plant is permitted by Texas' environmental regulator to emit particulate matter, carbon monoxide, nitrogen oxides and other pollutants – although not at very large volumes."
Sourcing: ~ partially sourced (counts as unsourced) — regulator named only generically, no permit or figures cited
Missing Content: The specific permit and its emission volume limits
Tactics: NLP — loaded qualifier ("although not at very large volumes" reassures without figures — mirrors Claim 3's water framing, applied to air)
Legal Exposure: low — vague permit characterization without the underlying document
Recommendation: Seek source

Claim #17

Text: "'Our facilities comply with all federal, state and local air quality and primary requirements, which are designed to protect public health,' Cosic said."
Sourcing: ✓ sourced — direct, named, on-record quotation
Tactics: Sycophancy / Authority Deference — answers a volume question with a compliance statement, implicitly equating compliance with acceptable impact rather than addressing the magnitude asked about
Legal Exposure: none
Recommendation: Publish with context

Claim #18

Text: "The plan said that the El Paso region emits 7.3 million tons of greenhouse gases annually."
Sourcing: ✓ sourced — City of El Paso Climate Action Plan (named document)
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #19

Text: "by 2050, the city's target is to reduce those emissions by 71% compared with current levels."
Sourcing: ✓ sourced — City of El Paso Climate Action Plan (named document)
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #20

Text: "So, if the McCloud power plant emits hundreds of thousands of tons of carbon dioxide annually, the city may be hard-pressed to meet its own climate and energy goals it spent multiple years and millions of dollars sketching out."
Sourcing: ✗ unsourced — reporter's own inferential construction, not attributed to the Climate Action Plan or any named source
Missing Content: No citation for the CO2 magnitude or the planning-cost figure
Tactics: NLP — causal framing ("So, if..." borrows credibility from the two sourced figures immediately preceding it); Logical Fallacy — unproven causation; Emotional Manipulation — loaded language ("sketching out" diminishes a cited multi-year planning process)
Legal Exposure: low — misleading-statistic risk, mitigated by the explicit conditional framing
Recommendation: Rephrase — likely the single sentence in the piece most at risk of being read as established fact by a skimming reader

Claim #21

Text: "Boyar Trejo is proposing a community benefit agreement between the city and Meta Platforms Inc."
Sourcing: ✓ sourced — reporter's direct account; contents reproduced in detail later
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #22

Text: "Boyar Trejo last week voted against terminating the city's economic development agreement with Meta that was approved in late 2023, citing concerns that the city could face massive legal damages."
Sourcing: ✓ sourced — public record (City Council vote)
Tactics: None
Legal Exposure: none
Recommendation: Publish

Claim #23

Text: "Futureland neighborhood: Fund sewer infrastructure for 214 homes about one mile from the data center."
Sourcing: ✗ unsourced — no citation for the home count or distance
Tactics: None
Legal Exposure: none
Recommendation: Seek source

Claim #24

Text: "(The current Chapter 380 agreement requires 50 employees.)"
Sourcing: ✗ unsourced — no link or reference to the agreement text
Tactics: None
Legal Exposure: low — inaccurate-legal-characterization risk (specific numeric term of an agreement, stated without citation)
Recommendation: Seek source

Claim #25

Text: "the expanded job projection to 300 permanent roles"
Sourcing: ✗ unsourced — no citation for the figure
Tactics: None
Legal Exposure: none
Recommendation: Seek source

Claim #26

Text: "The proposal by Boyar Trejo comes after a highly contentious daylong City Council meeting June 9 in which the council in a 5-3 vote declined to terminate the tax break agreement with Meta. Over 180 public speakers called to end the deal even in the face of potential litigation, with numerous tradesmen and construction workers and others supporting the deal."
Sourcing: ✓ sourced — public record (council vote) plus reporter's direct observation/count
Tactics: None
Legal Exposure: none
Recommendation: Publish

High-Risk Flags

None. No claim was rated high legal exposure.

Unsourced Claims

#7, #10, #11, #13, #14, #16, #20, #23, #24, #25 (10 of 26 — see per-claim detail above for reasons)

Per-Claim Tactic Summary

• Sycophancy / Authority Deference: 1 instance (#17)
• Emotional Manipulation: 1 instance (#20)
• Logical Fallacies: 4 instances (#7 false equivalence, #10 false precision, #11 false precision, #20 unproven causation)
• NLP / Linguistic Persuasion: 4 instances (#3 presupposition, #13 nominalization, #16 loaded qualifier, #20 causal framing)

Document-Level Tactic Analysis

Fifth-gen warfare patterns: None detected at a level warranting a formal flag. No memetic worldview installation, narrative hijacking, asymmetric escalation, information fragmentation, or accusation-in-a-mirror. The community-vs-corporation tension present is a property of the underlying dispute (data center siting), not a reporting choice — the piece structurally balances critic, company, and technocratic-utility voices.

Meta-manipulation: Not detected. This is a straight discussion recap, not a document that teaches or warns against manipulation while itself deploying it.

Quoted-rhetoric observations (outside the formal claim list, worth noting): Crouse's rhetorical question — "Has any kind of health studies been done on the impacts, long-term exposure of data centers?" — presupposes an unstudied risk without asserting one; it's a legitimate question but functions rhetorically to plant concern. Trevizo's "just yelling passionate things back and forth" mildly frames the opposition's approach as unproductive/emotional. Both are ordinary quoted stakeholder rhetoric, not reporter-driven tactics, and are noted here rather than tagged to a numbered claim since neither contains an independently verifiable factual assertion.

Connective Tissue NLP Summary

Opening sentence (highest-leverage embedding point): "A roundtable discussion Thursday involving El Paso city officials, Meta Platforms Inc. and critics of the company's incoming Northeast data center highlighted one thing: the data center's impacts are too myriad and complicated to hash out in 60 minutes." To accept this framing, the reader must grant that the meeting's core finding was its own inadequacy of format — a complexity claim that pre-empts any single answer in the piece from counting as resolution. This primes the reader to expect, and forgive, the absence of firm answers on water, air, and jobs that follows. Defensible editorial judgment for a discussion recap, but still an embedded frame accepted before a single fact appears.

Transition phrases:

• "So, if the McCloud power plant emits hundreds of thousands of tons..." — "So" presents an unsourced CO2 estimate as the logical consequence of the two sourced Climate Action Plan figures immediately before it (see Claim 20).
• "Beyond hosting talks, Boyar Trejo is proposing..." — additive, not causal; flagged only by contrast with the "So" construction above. Most transitions in this piece are neutral sequencing.

Qualifiers:

• "not as large as it is being portrayed" (Claim 3) — presupposes an unnamed exaggerated portrayal.
• "although not at very large volumes" (Claim 16) — same reassurance move, applied to air emissions.
• "millions of dollars sketching out" (Claim 20) — diminishing framing of the city's own cited planning process.

Sequencing: critic grievance → official reassurance (water) → official reassurance (air) → sourced Climate Action Plan stat → unsourced "So, if" extrapolation → mayor's concern quote → councilmember's benefit-agreement proposal. The arc moves from unresolved critic questions toward increasingly concrete institutional response, ending on Boyar Trejo's detailed policy proposal — which reads as the piece's implicit answer to the concerns raised earlier, even though the meeting itself (per the reporter's own opening framing) resolved nothing.

Term consistency: "the data center" (Meta's) vs. "the McCloud facility"/"McCloud power plant" (El Paso Electric's separate gas plant) are kept distinct throughout — a precision worth noting, since the two are topically adjacent and easy to conflate downstream. "Restoration" is used consistently by Meta as a term-of-art for its water-offset program (not remediation) — the Tavistock pre-screen flagged 4 repetitions as a pairing move, but on review this is consistent term-of-art usage, not manipulative reframing.

Recommendation Summary