Threshold-Trigger Elision and Carve-Out Inversion in the CFTC's 2024 amendments to Regulation 1.25. Two frontier AI models tested by the RLB Specialist Panel produced confident, fluent reconstructions of the rule that dropped the size triggers governing the 50 per cent concentration ceiling, inverted the DWAM carve-out set, returned a no-standard answer for direct U.S. Treasury obligations, and drifted from the March 31, 2025 SIDR compliance anchor.

The audit covers concentration-limit recital, DWAM portfolio testing, and SIDR Report compliance scheduling. The audience exposed is lawyers who use frontier AI assistants for fund-formation memoranda, investment policy statement reviews, partner-level briefings, and client-facing compliance calendars.

The pattern in one line

Frontier AI models reading the CFTC's 2024 amendments to Regulation 1.25 dropped the dual size triggers that activate the 50 per cent concentration ceiling, substituted U.S. Treasury repurchase agreements into the DWAM exclusion set in place of the regulator's three carved-out classes, returned a no-DWAM-standard answer for direct U.S. Treasury obligations, and drifted from the March 31, 2025 SIDR compliance anchor into a relative formulation. For lawyers the consequence lands in the fund-formation memoranda the audience produces from AI-assisted drafting, the failure is operationally silent because the model output reads as an adjudicated answer rather than a dropped-axis one.

How the RLB Specialist Panel tested this

The RLB Specialist Panel designed audited questions to mirror how lawyers actually use frontier AI assistants on Regulation 1.25: drafting fund-formation memoranda, investment policy statement reviews, partner-level briefings, and client-facing compliance calendars, scoping concentration-testing controls against the 50 per cent ceiling, building DWAM portfolio testing against the 24-month standard, and scheduling SIDR Report compliance against the regulator's published anchor. Questions are prepared by the RLB Specialist Panel based on real practical AI usage in the workflows the respective audience uses AI for. The Panel binds each AI finding to verbatim regulator-issued source text held as primary substrate.

The substrate sources cover 17 CFR 1.25(b)(3)(ii) (the size-triggered concentration ceiling), 17 CFR 1.25(b)(3)(iv) (the 24-month portfolio DWAM standard and its carve-out set), and the operative section of the Commodity Exchange Act at 7 USC 6d. Substrate preparation is internal to the RLB Specialist Panel. The audited subjects were two frontier AI models, run with web search enabled, against Specialist Panel application-style questions.

What the models got wrong

Finding 1: size-trigger elision on the 50 per cent concentration ceiling (Claude Opus 4.7). Claude Opus 4.7 wrote that "the final rule did not adopt tiered or FCM-size-based thresholds, the percentage limits apply uniformly regardless of FCM size." The regulator's text at 17 CFR 1.25(b)(3)(ii) keys the 50 per cent ceiling to two thresholds simultaneously: the fund's own assets at one billion dollars and the management company's assets at twenty-five billion dollars.

The model surfaced the FCM-size axis correctly (the rule is not keyed to FCM size) while dropping the fund-side and management-company-side triggers that actually govern when the ceiling engages. The Panel records this as inference_drift, bound to the regulator's verbatim text and published with citation RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q001-Opus47.

Finding 2: size-trigger elision plus a fabricated FCM-side tier label (Claude Sonnet 4.6). Claude Sonnet 4.6 wrote that "there is no size-based tier that changes the percentages based on the FCM's total assets" and then offered a Tier 1 per-instrument 10 per cent ceiling for any single government money market fund. The negation reads as an adjudication of the size question, but it drops both axes the regulator does key on (fund asset size, management-company asset size) and adds a tier label that does not appear in the rule.

A risk officer reading the answer would walk away believing the size question has been resolved. The Panel records this as inference_drift, bound to the regulator's verbatim text and published with citation RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q001-Sonnet46.

Finding 3: DWAM carve-out inversion: Treasury repos substituted for the actual exclusion set (Claude Opus 4.7). Asked which asset classes are excluded from the portfolio-level 24-month dollar-weighted average maturity calculation, Claude Opus 4.7 wrote that "U.S. Treasuries held under repurchase agreements" are excluded. The regulator's carve-out set under 17 CFR 1.25(b)(3)(iv) is government money market funds, Treasury ETFs, and foreign sovereign debt. U.S. Treasury repos are not on the exclusion list. The model swapped one adjacent asset class into the carve-out and dropped the three classes the regulator actually excludes.

The Panel records this as inference_drift, bound to the regulator's verbatim text and published with citation RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q002-Opus47.

Finding 4: no-DWAM-standard answer for direct U.S. Treasury obligations (Claude Sonnet 4.6). Asked whether a portfolio maturity constraint applies to direct U.S. Treasury securities, Claude Sonnet 4.6 concluded that "no DWAM standard or individual-maturity cap found in the 2024 amendments applies to that category." The regulator's 24-month portfolio DWAM standard applies to direct U.S. Treasury obligations by default; the carve-out set covers government money market funds, Treasury ETFs, and foreign sovereign debt, not direct Treasuries.

The model returned a no-standard answer to a question the regulator answers with a standard, the most operationally consequential shape in the audit because it tells the reader that no compliance work is required where the regulator requires it. The Panel records this as inference_drift, bound to the regulator's verbatim text and published with citation RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q002-Sonnet46.

Finding 5: drift from the March 31, 2025 SIDR compliance anchor (Claude Opus 4.7). Asked for the separate compliance deadline for updating the Segregation Investment Detail Report and customer risk disclosure statement, Claude Opus 4.7 anchored the deadline at "a separate, later date (commonly described as roughly six months to a year after the effective date)." The regulator's published anchor is March 31, 2025. The model retrieved the structural distinction (SIDR has its own date separate from the general effective date) and then drifted on the specific anchor, returning a generic range where the regulator records a date certain.

The Panel records this as inference_drift, bound to the regulator's verbatim text and published with citation RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q004-Opus47.

The combined pattern across the five findings: one axis of a multi-trigger condition is surfaced correctly while the other axes are dropped, narrow exclusion sets are substituted with superficially-similar adjacent classes, and date-certain compliance anchors are replaced with generic relative ranges.

Why this matters for lawyers

A partner-level memorandum that recites the 50 per cent ceiling as a uniform FCM-size-independent limit would misclassify the size-trigger structure of the rule. A client compliance calendar that anchors the SIDR update at "six months to a year after the effective date" would miss the regulator's March 31, 2025 date by an unbounded margin. A DWAM clause drafted around U.S. Treasury repos as a carved-out class would exclude the wrong book from concentration testing and over-include the right ones.

The five findings together touch every operative axis of the 2024 amendment package: the size-triggered 50 per cent concentration ceiling, the 24-month DWAM portfolio standard and its three-class carve-out, and the March 31, 2025 SIDR and customer risk disclosure compliance anchor. None of these are recoverable from the AI output alone. The model output reads as a confident reading of the amended rule, and the reader cannot validate it without independently knowing the dual-trigger structure of 17 CFR 1.25(b)(3)(ii), the three carved-out classes under 17 CFR 1.25(b)(3)(iv), and the date certain of March 31, 2025.

The regulator's actual position

The Commodity Futures Trading Commission, in the 2024 amendments to Regulation 1.25, sets out the following operative positions, each recorded in verbatim regulator-issued source text held by the RLB Specialist Panel as primary substrate:

• Concentration ceiling (17 CFR 1.25(b)(3)(ii)): Investments in government money market funds or qualified ETFs whose assets are at least one billion dollars and whose management company manages at least twenty-five billion dollars may not exceed 50 per cent of total segregated assets. The ceiling is conditional and engages only when both size triggers are simultaneously satisfied.
• DWAM portfolio standard (17 CFR 1.25(b)(3)(iv)): The dollar-weighted average of the time-to-maturity of the portfolio may not exceed 24 months. The carve-out set, the asset classes excluded from the calculation, is government money market funds, Treasury ETFs, and foreign sovereign debt.
• Direct Treasury treatment: Direct U.S. Treasury obligations and U.S. Treasury repurchase agreements are not on the carve-out list. The 24-month portfolio DWAM standard applies to them by default.
• SIDR and customer risk disclosure compliance anchor: SIDR Report and customer Risk Disclosure Statement updates carry a compliance date of March 31, 2025, separate from the general effective date for conforming investment policies.

The five published findings each misstate at least one of these positions in ways the regulator's own text resolves.

What this tells us about AI for lawyers

The findings indicate that frontier AI assistants used in lawyers workflows on customer-funds investment policy work cannot be relied on for: verbatim recital of multi-axis size-trigger conditions on the 50 per cent ceiling; verbatim recital of narrow carve-out sets under the DWAM standard; default-application answers on asset classes the standard covers by default; or date-certain reproduction of separate compliance anchors such as the March 31, 2025 SIDR date. Each of these question types is handled by the AI in a confident register that masks the dropped axis, the substituted carve-out class, the no-standard answer, or the relative-date drift.

The implication for lawyers is that AI-assisted Regulation 1.25 reconstructions are drafting starts and not citation-grade outputs; the fund-formation memoranda must be matched against the regulator's published text before reliance.

What the RLB Specialist Panel is doing about it

The RLB Specialist Panel records confirmed hallucination findings under a methodology that requires a verbatim regulator excerpt for every documented claim. The five findings on Regulation 1.25 are open-access on the Regulation 1.25 hub on RegLegBrief.com, with immutable RLB Citation IDs that lawyers can reference in client memos, control narratives, audit programmes, or board-level briefings without paywall, registration, or data-licensing constraints. The Panel records alongside each finding the verbatim AI subject answer, the verbatim regulator source text, the failure-mode classification, and the substrate document the regulator text is bound to.

AI labs and model developers named in any published finding have an unconditional right of reply. Audit, compliance, treasury, legal, risk, and operations teams interested in partnership with the RLB Specialist Panel on additional regulatory surfaces can reach the Panel via the same channel.

What lawyers teams should do

• Verify every concentration-ceiling recital against the verbatim text of 17 CFR 1.25(b)(3)(ii), including both size triggers (fund-asset side at one billion dollars, management-company side at twenty-five billion dollars).
• Match every DWAM clause against 17 CFR 1.25(b)(3)(iv) for the carve-out set: government money market funds, Treasury ETFs, and foreign sovereign debt. Direct U.S. Treasury obligations and Treasury repos are not in the carve-out.
• Anchor SIDR and customer risk disclosure compliance entries to the date certain of March 31, 2025, not to a relative formulation off the general effective date.
• Treat AI-assisted Regulation 1.25 summaries as drafting starts, not citation-grade outputs. Verbatim quotation of any threshold, carve-out, or compliance date must be sourced from the regulator's published text.

The full audit on Regulation 1.25 is at https://reglegbrief.com/regulators/j2/us/cftc/fcm-dco-customer-funds-investments-reg-1-25-2024/. The five published findings each carry an immutable RLB Citation ID and verbatim regulator-issued source text suitable for direct lawyers reference.

Right of Reply

These findings and associated work have been put up in public with a view of the greater good for the development of a safer AI ecosystem. Any party reading this or any finding on reglegbrief.com may contact us and have an unconditional right of reply; the Specialist Panel will publish any factual correction or contextual response alongside the original finding, with no editorial gatekeeping. Researchers, regulators, and compliance teams with questions on methodology or specific findings can reach the Specialist Panel via the same channel.

Source & Methodology Standards

RegLeg Brief is operated by Verdus Technologies Pte. Ltd. (UEN 201616982R), incorporated in Singapore. The RLB Specialist Panel, with an aggregate of over 60 years of public-policy and industry experience, documents only confirmed hallucination findings, under a methodology that requires a verbatim regulator excerpt for every documented claim. All findings, citation IDs, model outputs, regulator excerpts, and methodology notes are open-access.

Primary source verified: CFTC Amendments to Regulation 1.25, Permissible Investments of Customer Funds by Futures Commission Merchants and Derivatives Clearing Organizations (2024) · Substrate documents: R1-ACT-Q4_7_USC_6d_Commodity_Exchange_Act.pdf, p_03_REGULATION_17_CFR_1_25_b__3__ii____two_tier_asset_b_download.pdf, p_03_REGULATION_17_CFR_1_25_b__3__iv____dollar_weighted_download.pdf · eCFR: ecfr.gov · CFTC: cftc.gov

Citation IDs referenced:

• RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q001-Opus47
• RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q001-Sonnet46
• RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q002-Opus47
• RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q002-Sonnet46
• RLB-H-US-CFTC-FCM-DCO-CUSTOMER-FUNDS-INVESTMENTS-REG-1-25-2024-Q004-Opus47