Frontier AI models over-generalised a phased pilot sunset, fabricated a staff FAQ, and dropped the OCC interpretive hook on the CFTC's digital asset margin framework.

Two frontier AI subjects tested by the RLB Specialist Panel inverted the post-onboarding reporting cadence, fabricated a March 2026 CFTC Staff FAQs source, dropped the OCC Interpretive Letter 1183 cross-reference, and defaulted to the base 20 per cent haircut where the regulator's worst-case selection rule governs.

The pattern in one line

Frontier AI models tested on the CFTC Digital Asset Collateral Framework returned answers that read as authoritative, but inverted the post-onboarding weekly reporting cadence, omitted the OCC Interpretive Letter 1183 cross-reference that anchors national-trust-bank payment stablecoin eligibility, and replaced the multi-DCO worst-case haircut with the base floor.

How the RLB Specialist Panel tested this

The questions in this cell were prepared by the RLB Specialist Panel based on real, practical AI usage in the workflows that lawyers actually use AI for under the CFTC Digital Asset Collateral Framework. Each question targets a specific deliverable type where an AI assistant is plausibly the first draft: a memo, an eligibility paragraph, an onboarding checklist line, a haircut-model assumption, a regulator-facing filing sentence. The Panel issued each question to two frontier AI subjects with web search active.

The Panel then bound every AI response to verbatim regulator-issued source text held as primary substrate, comparing the model output against the CFTC staff letter text and the regulator-issued source documentation for each provision. Only responses where the AI subject was demonstrably wrong against the verbatim regulator-issued source text are published as findings; responses that were substantively correct, or that refused on calibration grounds, are retained internally and not surfaced.

What the models got wrong

Finding: Payment stablecoin eligibility memo dropped the OCC Interpretive Letter 1183 cross-reference. The Specialist Panel asked, in application form, which CFTC staff letter is operative for FCM acceptance of payment stablecoins backed by reserves at an OCC-chartered national trust bank, what amendment to the issuer definition the operative letter introduced, and what specific regulatory instrument provides the eligibility hook for national-trust-bank issuers.

Claude Opus 4.7 with web search active answered that Staff Letter 25-40 was reissued as Staff Letter 26-05 on February 6, 2026, with a limited revision expanding the payment stablecoin definition to include stablecoins issued by national trust banks (RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q005-Opus47). Claude Sonnet 4.6 with web search active reached the same substantive description of the amendment, characterising the change as a targeted amendment to the payment stablecoin definition to add OCC-chartered national trust banks as a permitted issuer category (RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q005-Sonnet46).

The substrate held by the Panel records the operative regulator-issued instrument: Staff Letter 26-05 was issued as a limited revision to 25-40 expanding the payment stablecoin definition to include national trust banks, and the eligibility hook for those issuers sits in OCC Interpretive Letter 1183, which both model outputs either omitted entirely or surfaced only as a downstream reference. The substantive description of the amendment is right; the cross-document interpretive hook that anchors national-trust-bank eligibility in federal interpretive authority is missing.

Finding: Weekly digital asset reporting obligation classified as sunsetting at month four, when the regulator continues it. The Specialist Panel asked, in application form, which onboarding conditions cease at the end of the initial three-month phase and which continue beyond it for an FCM accepting bitcoin, ether, and USDC as customer margin collateral. Claude Opus 4.7 with web search active answered that weekly reporting of total digital assets held in customer accounts by asset type for each of the three customer account classes terminates at the end of the initial three-month phase under Letter 25-40 (RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q006-Opus47).

Claude Sonnet 4.6 with web search active reached the same end-state and added a fabricated authority: the model cited March 2026 CFTC Staff FAQs as the procedural source confirming that weekly digital asset reporting ceases at the end of the third calendar month following the firm's notice filing (RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q006-Sonnet46). The substrate held by the Panel records the regulator's actual position: after the initial 3-month phase, asset-type restrictions and incident-reporting conditions no longer apply, but weekly reports of total crypto assets held in each of the futures, foreign futures, and cleared swaps customer accounts continue.

There is no March 2026 CFTC Staff FAQs document of the kind Sonnet 4.6 cited. This finding combines an inverted obligation classification with a fabricated source citation.

Finding: Multi-DCO haircut tiebreaker rule replaced by the base 20 per cent floor. The Specialist Panel asked, in application form, how the customer margin haircut for digital assets is calculated when multiple registered DCOs each accept the same digital asset but at different haircut rates, a scenario that arises whenever an FCM clears the same crypto asset through more than one DCO, which is the commercially dominant pattern for bitcoin, ether, and the eligible payment stablecoins.

Claude Sonnet 4.6 with web search active answered that for digital assets not accepted by any registered DCO as initial margin, the FCM must apply a minimum 20 per cent haircut to the current market value of the customer-deposited collateral, and presented this base-case framing as the operative rule for the multi-DCO scenario the Panel asked about (RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46). The model's substantive description of the 20 per cent floor is correct as far as it goes: the staff letter does set a minimum 20 per cent haircut for digital assets that no registered DCO accepts as initial margin.

The model's error is in transplanting that base-case rule onto the multi-DCO conditional. The substrate held by the Panel records the regulator's actual selection rule: where multiple DCOs accept the same asset at different haircuts, the FCM must apply the highest such haircut. The 20 per cent floor the model described is the regulator's base case for the no-DCO-accepts-it scenario; it is not the operative rule for the multi-DCO scenario, where the worst-case selection rule governs.

A customer margin programme that operationalises the model's framing systematically under-collateralises customer accounts that hold the same digital asset across two or more DCOs at differing haircut rates, which is the realistic operating environment rather than the edge case.

Why this matters for Lawyers

Lawyers advising on the CFTC Digital Asset Collateral Framework operate in a high-citation-fidelity environment. Eligibility opinions, FCM onboarding memos, customer collateral due-diligence, and regulator-facing filings are all read line-by-line by clients, counterparties, regulatory reviewers, and supervisors. The Q006 weekly reporting inversion is the more serious failure: a legal opinion structured on a weekly-reporting-sunsets framing produces wrong client guidance, wrong onboarding documentation, and exposes the FCM to an ongoing reporting violation that accrues from the post-onboarding date.

The Q005 payment stablecoin OCC 1183 omission and the Q007 multi-DCO haircut substitution are less directionally wrong but equally fatal to the credibility and defensibility of the legal advice carrying them. Each error is identifiable on first review of the deposited staff letter, and the wider opinion loses credibility once the reader spots a verifiable defect.

The regulator's actual position

Staff Letter 26-05 with OCC Interpretive Letter 1183 as the eligibility hook. Staff Letter 26-05 is the operative CFTC instrument for FCM acceptance of payment stablecoins backed by reserves held at an OCC-chartered national trust bank. The amendment expands the payment stablecoin issuer definition to include national trust banks. The interpretive authority that grounds national-trust-bank eligibility, however, sits outside the staff letter itself: it is OCC Interpretive Letter 1183, which the CFTC staff letter cross-references and on which the eligibility analysis depends.

Any payment stablecoin eligibility memo, due-diligence representation, or counterparty disclosure that omits the OCC 1183 hook is incomplete on the controlling interpretive chain.

Partial sunset at month four; weekly reporting continues. Staff Letter 25-40, as amended by 26-05, sets the conditions that terminate at the end of the initial three-month onboarding phase and the conditions that continue. The verbatim regulator-issued formulation records the partial nature of the sunset: asset-type restrictions and incident-reporting conditions no longer apply after the initial 3-month phase, but weekly reports of total crypto assets held in each of the futures, foreign futures, and cleared swaps customer accounts continue. The continuing weekly reporting obligation is not among the conditions that sunset.

There is no procedural CFTC Staff FAQs document keyed to March 2026 that terminates the weekly cadence.

Multi-DCO scenario: highest accepted haircut governs. The CFTC staff letter contains two distinct 20 per cent figures and they govern different scenarios. One is a haircut floor for digital assets that no registered DCO accepts as initial margin: in that case, a minimum 20 per cent haircut to current market value applies to customer-deposited collateral. The other 20 per cent applies to a separate proprietary-holdings concentration limit, a distinct rule on the firm's own book rather than the customer book. Neither of these is the operative rule for the multi-DCO conditional.

For the multi-DCO scenario where two or more registered DCOs each accept the same digital asset at different haircut rates, the operative selection rule is the highest such haircut, applied to the customer's posted collateral. The structural logic is the regulator's worst-case selection: where a market-recognised haircut range exists, the FCM applies the top of the range, not an arbitrary floor and not the lowest-DCO rate. The base 20 per cent floor is the rule for the no-DCO-acceptance case; it is not the rule for the multi-DCO case.

What this tells us about AI for Lawyers

There are two distinct lessons here for lawyers working with AI on the CFTC Digital Asset Collateral Framework. The first, on the weekly digital asset reporting obligation, is the more serious. The AI subjects' answers are not surface citation slips: they invert the direction of an ongoing reporting obligation. Where the regulator's text says the weekly cadence continues, the AI subjects said it sunsets. One subject went further and cited a fabricated March 2026 CFTC Staff FAQs source to support the inversion.

A reader who runs an AI-drafted onboarding memo through a routine citation-check workflow will not catch this; the staff-letter reference cited is the actual operative letter. Only a substantive read of the post-phase obligation text catches the inversion. This is the failure mode AI Labs need to know about, and it is the failure mode practitioners need to defend against most rigorously.

The second, on the payment stablecoin OCC 1183 hook and the multi-DCO haircut tiebreaker, is the more familiar pattern: the substantive paraphrase is close enough that the answer reads as authoritative, but a load-bearing qualifier or cross-reference has been silently dropped. Citation-checking workflows will catch this only if they run against the full staff-letter text and the cross-referenced interpretive instruments, not against a chained reference.

The defensive posture is the same in both cases: always anchor the citation against the operative CFTC staff letter and its cross-referenced authorities, not against a secondary commentary, before letting an AI-drafted output go out the door.

What the RLB Specialist Panel is doing about it

The RLB Specialist Panel is engaging with the AI subjects' developers and with practitioner audiences working under the CFTC Digital Asset Collateral Framework. The Panel maintains an audit register of confirmed hallucinations bound to verbatim regulator-issued source text, surfaces them on the live regulation page and on each audience-specific briefing, and accepts right-of-reply submissions from the AI subjects' developers and from regulator-side reviewers.

For lawyers this means the same questions can be re-issued against successor model releases; the bound substrate makes it straightforward to verify whether a specific failure mode has been corrected upstream, or whether the same hallucination is still being produced. Partnership briefings with AI labs are offered against the audit register, not against synthesised demonstrations, so the corrections that matter are evidenced against the operative staff letter text rather than against a paraphrase chain.

What Lawyers teams should do

For lawyers drawing on AI in workflows that touch the CFTC Digital Asset Collateral Framework, the practical action items are direct:

• Anchor every CFTC Digital Asset Collateral Framework citation produced by an AI assistant against the operative CFTC staff letter text and its cross-referenced interpretive authorities, before the citation enters a memo, opinion, or contractual representation.
• Treat any AI-generated statement on the post-onboarding obligation set as a high-risk paragraph: read the operative staff letter directly and confirm that weekly digital asset reporting continues beyond month four before relying on the AI characterisation.
• When advising on payment stablecoin eligibility, validate that the OCC Interpretive Letter 1183 cross-reference is surfaced in the memo, not just the CFTC staff letter amendment.
• When advising on customer margin haircut methodology, validate that the multi-DCO highest-accepted-rate selection rule is reflected in the memo, not the base 20 per cent floor.
• Maintain a firm-level AI assist log that records every AI-drafted CFTC staff-letter citation entering a client deliverable, so that downstream verification and post-hoc audit are traceable.
• Subscribe to the RLB Specialist Panel audit register on the CFTC Digital Asset Collateral Framework to receive notifications when a confirmed AI hallucination is added, withdrawn, or corrected against successor model releases.

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 Staff Advisory on Digital Asset Collateral and Tokenized Assets (2025) · Substrate documents: p_02_GUIDELINE_CFTC_Staff_Letter_25_40___26_05___two_di_download.pdf, p_02_GUIDELINE_CFTC_Staff_Letters_25_40___26_05___post_download.pdf, p_03_NOTICE_CFTC_Staff_Letter_26_05__February_6__202_download.pdf · CFTC: cftc.gov

Citation IDs referenced:

• RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q005-Opus47
• RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q006-Opus47
• RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46