Source-Credit Fabrication on CCP override framework disclosure. A frontier AI model, asked what CCPs must publicly disclose about their margin model override framework under the CPMI-IOSCO Initial Margin Disclosure Consultation, returned a confident three-part enumeration that does not appear in the consultative document and framed an expectation as a mandatory obligation. For a company secretary drafting a board paper, the misframing is the kind that becomes a board-level commitment before anyone looks at the primary regulator text.

The pattern in one line

A frontier AI assistant, asked a precise question about the public disclosure scope for a CCP's margin model override framework, returns a structured three-element specification, frames the obligation as mandatory, and supports the response with a secondary commentary URL rather than the primary BIS text.

How the RLB Specialist Panel tested this

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. For this finding, the Panel posed a Specialist Panel application-style question to two frontier AI subjects, framing the request the way a company secretary or board paper drafter would type it into an AI assistant when preparing the next CCP-counterparty risk update for the audit committee or the board risk committee.

The Panel then bound the model output against the verbatim consultation text on CCP override framework disclosure held as primary substrate, and against the broader d232 cover note on CCP margin model overrides.

The Panel does not give the AI subjects the substrate. The Panel evaluates only the AI subject's own output against the substrate the Panel holds. That asymmetry is what reveals the failure: it shows the model output a company secretary will actually receive when the company secretary is not yet looking at the BIS text.

What the models got wrong

Claude Sonnet 4.6, web search on, asked what specific information CCPs must publicly disclose about their margin model override framework under the 2026 consultation on initial margin transparency, returned a three-part enumeration: (1) the instances or circumstances where overrides may be warranted; (2) the key decision-makers authorised to exercise override discretion; and (3) the permissible types of adjustments that can be made.

None of those three categories appears in the consultation. The consultation states: "CCPs should publicly disclose relevant information on their override framework." The specific content of "relevant information" is left open for comment. The model framed the disclosure as a "must", not a "should". It returned a closed three-part list, not an open expectation. It supported the answer with a secondary commentary URL, not the primary BIS d232 cover note.

A board paper that adopts the three-part enumeration as the disclosure scope description records, on the board minutes, that the entity is taking a position on what the consultation requires. That position is not in the consultation.

Citation: RLB-H-INT-BIS-CPMI-IOSCO-INITIAL-MARGIN-DISCLOSURE-CONSULT-2026-Q005-Sonnet46.

Why this matters for company secretaries

The board paper is the entity's documented understanding of the obligation. The minute is the record of the board's adoption of that understanding. If both are structured around a fabricated three-part standard, the entity has, on its own records, committed to a disclosure framework that the regulator did not require.

When the regulator examines the entity's CCP counterparty governance, or when the entity files a comment-letter submission asserting the three-part standard on behalf of its operating entity, the file shows an entity that built its position on a specification not derived from the source document. The company secretary's review is the last documented checkpoint before the board paper goes into circulation. The audit trail of that review becomes evidence in any later examination.

For an internationally active investment bank that holds CCP counterparty exposures across multiple clearinghouses, the board paper also drives the credit risk and collateral management committee's view of which CCPs are "compliant" with the disclosure framework. A compliant CCP that does not enumerate the three fabricated categories will be flagged as deficient against the fabricated standard, triggering corresponding internal remediation activity with no regulatory basis.

The regulator's actual position

The consultative document states, verbatim: "CCPs should publicly disclose relevant information on their override framework." The cover note for d232 sets the override framework disclosure as one of several areas in which CCPs are expected to provide relevant information; it does not enumerate the contents of that disclosure, and it does not frame the obligation as a hard requirement. The decision to leave the specific content open for comment is itself a substantive feature of the consultation.

The CPMI-IOSCO Secretariat's standard formulation is "should", not "must". The Secretariat reserves the prescriptive formulation for finalised standards. The drift from "should" to "must" is, in regulatory terms, the difference between an expectation that anchors supervisory dialogue and a binding line-item rule.

What this tells us about AI for company secretaries

The failure class is Source-Credit Fabrication. The model attaches a confident, structured factual claim to a regulator source that does not contain the claim, and supports the attribution with a third-party commentary URL. For company secretaries, the operational consequence is direct: any AI output reporting "the regulator requires (i), (ii), and (iii)" must be verified against the cited regulator text, and the citation itself must be verified against the primary source.

The second signal is that AI output framing a consultation as a "must" is presumptively drifted. Consultations almost always use expectation language. The drift is small in word count and large in board-paper effect. The company secretary's pre-circulation review must catch it every time.

The third signal is that a structured three-part enumeration looks more like a settled standard than a one-line expectation. The structure itself, not just the content, conveys settledness. A board reviewing the paper reads the structure as a regulator-issued specification. The board secretary's role is to ensure the structure in the paper matches the structure in the source.

What the RLB Specialist Panel is doing about it

The Panel issues the finding bound to the verbatim consultation text held as primary substrate, with the citation ID assigned for cross-reference. The Panel does not assess the model's overall accuracy on the consultation. The Panel assesses the specific output against the specific regulator text on the specific question a board paper drafter is likely to ask.

For company secretaries, the finding is available as a worked example of the verification discipline required when AI is used in board paper drafting on regulator-issued source text. For audit committee chairs and board risk committee chairs, the finding is available as a worked example of the kind of failure mode the entity's verification controls must catch before the paper reaches the board.

Citation: RLB-H-INT-BIS-CPMI-IOSCO-INITIAL-MARGIN-DISCLOSURE-CONSULT-2026-Q005-Sonnet46.

What company secretaries teams should do

Action items for a company secretary's office supporting boards of CCPs, clearing members, or internationally active investment banks on CCP disclosure governance:

• Treat any AI-returned three-part enumeration of regulator-issued obligations as a presumptively unverified claim: verify the enumeration line by line against the primary regulator text before the paper enters the board pack.
• Reject AI output that frames a consultation as imposing a "must" requirement: the CPMI-IOSCO Secretariat's standard formulation on consultation-stage text is "should", and any drift to "must" is the most common AI drift on this document class.
• Require the drafter of any board paper on CCP disclosure to identify, against the primary regulator text, the verbatim source for each obligation cited in the paper, and to record the verification in the paper's audit trail.
• Reject secondary commentary URLs as citation sources for board papers: if the AI cites a third-party commentary rather than the primary BIS text, treat the underlying factual claim as unverified.
• Brief audit committee and board risk committee chairs on the Source-Credit Fabrication failure class, so the committee chair has the vocabulary to challenge structured enumerations in papers that do not match the source.
• Use the RLB Specialist Panel's published findings as a working library of AI failure modes on regulator text, and use the failure class as the standard reference in the entity's AI-use policy for governance work product.

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: CPMI-IOSCO Consultative Report d232, Streamlining Variation Margin Disclosure (2026) · Substrate documents: p_03_ANNEX_Override_framework___public_disclosure_r_d232_covernote.pdf · CPMI portal: bis.org/cpmi

Citation IDs referenced:

• RLB-H-INT-BIS-CPMI-IOSCO-INITIAL-MARGIN-DISCLOSURE-CONSULT-2026-Q005-Sonnet46