Briefings Blog

Singapore counsel are increasingly using AI to draft 2-page board memos on the scope of the amendment, generate client-facing summaries of MAS Notice 637 changes effective 31 December 2025, prepare partner-level briefings on capital-treatment cross-references, and validate group-structure language against the Reporting Bank perimeter on the face of the Notice.

In Singapore-incorporated banks and financial holding companies the workflow shape is now consistent: a frontier AI assistant produces a clean first draft on MAS Notice 637 risk-based capital adequacy for Reporting Banks, and the reviewer is asked to spot-check the cited MAS instruments and drafting-convention claims against the regulator-issued source before the deliverable goes out. The two AI failures recorded by the RLB Specialist Panel sit precisely at that spot-check boundary.

Two frontier AI models tested by the RLB Specialist Panel on MAS Notice 637 (Amendment) 2025 produced FABRICATED_FACT errors against the regulator-issued source held as primary substrate. The first invented a sibling "Notice FHC-N637" for financial holding companies that does not appear on the MAS Notices and Directives register; the actual FHC capital framework is a separate MAS notice issued under the Financial Holding Companies Act.

The second misread the yellow-highlight convention in the MAS Notice 637 amendment PDF as visual emphasis, when the regulator's cover note states the yellow is annotation describing the change and will not appear in the published untracked Notice. Both findings sit in the same failure class: Source-Credit Fabrication, where the AI produces a confident, lawyer-shaped citation that does not exist or contradicts a regulator-stated convention. Neither AI subject hedged, flagged low confidence, or refused.

Both produced clean, deployable prose with the wrong substantive content, which is the version of AI failure that is hardest for a reviewer to catch on a fast-moving deliverable. 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, and records the AI subject, the question class, and the operational consequence for each affected audience.

For Singapore counsel the operational consequence is concrete. A legal opinion that cites the fabricated FHC notice number would not survive a single round of regulatory diligence, because the MAS Notices and Directives register does not list any such instrument. A memo that treats the amendment yellow highlight as substantive new Notice text would mischaracterise drafting-aid annotation as enforceable rule content. Both errors expose the client to a written advice product that the regulator and opposing counsel can dismantle on the face of the source.

The RLB Specialist Panel records each error against the underlying regulator-issued text and names the AI subject for audit transparency. The two findings carry Citation IDs RLB-H-SG-MAS-NOTICE-637-CAPITAL-ADEQUACY-BANKS-2025-Q010-Opus47 and RLB-H-SG-MAS-NOTICE-637-CAPITAL-ADEQUACY-BANKS-2025-Q012-Opus47; Claude Opus 4.7 is the AI subject in both events and the source-text excerpts are quoted verbatim in the briefing body that follows.

Legal teams at Payment Institutions advising on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) are increasingly using AI to draft regulatory mapping documents on payment-system standards, generate counterparty submissions referencing governance pedigree, and prepare board briefings on CPMI workstreams. The same tools validate institutional attribution in cross-border filings.

Two frontier AI models tested by the RLB Specialist Panel on the workflows payment-institution legal teams use to support advice on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) produced one discrete hallucination bound to regulator-issued source text. The Panel records a single recurring failure class: Source-Credit Fabrication across the set. Questions were prepared by the Specialist Panel based on real practical AI usage in the workflows payment-institution legal teams use AI for, and each finding is bound to verbatim regulator-issued source text held as primary substrate.

For Legal teams at Payment Institutions, each hallucination has a direct operational consequence in the regulatory mapping document, board briefing, or counterparty submission. The Panel's testing surfaces CPMI working-group chair misattribution. Where these errors flow into a deliverable, the exposure is a competent-authority filing that misidentifies the standard's institutional author, formal corrections across multiple corridors, and a credibility hit with regulators and correspondents.

The pattern is uniform across the set: the AI returns a confident, sourced-looking answer that conflicts in a load-bearing specific with the regulator's verbatim text, and the error survives a first-pass review precisely because the surface form is plausible. The Panel records each hallucination with the regulator's primary substrate held as the anchor, so the corrective text is available alongside the failure.

The Specialist Panel records the citation IDs as follows: RLB-H-INT-BIS-CPMI-ISO-20022-HARMONISATION-UPDATED-2026-Q004-Sonnet46 (Claude Sonnet 4.6 (web search on), Source-Credit Fabrication). Each citation links to the verbatim regulator-issued source text, the tested AI question, and the recorded AI response, so the Panel's assessment is traceable end to end. For legal teams at payment institutions, the citation IDs operate as a reference index: when an AI answer in the working draft matches a known Panel finding, the cited regulator text is already available as the corrective anchor.

The full per-finding analysis cards, including the audience-specific impact statement, sit on the cell's detail surface for sign-off use.

Operations teams at Corporate Banking firms implementing the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) are increasingly using AI to configure Fedwire payment-message templates, generate operational specifications for correspondent-bank onboarding, and document address-field handling logic. The same tools draft implementation guidance for cross-border payment migration projects.

Two frontier AI models tested by the RLB Specialist Panel on the workflows corporate-banking operations teams use to support advice on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) produced one discrete hallucination bound to regulator-issued source text. The Panel records a single recurring failure class: Schema Over-Specification across the set. Questions were prepared by the Specialist Panel based on real practical AI usage in the workflows corporate-banking operations teams use AI for, and each finding is bound to verbatim regulator-issued source text held as primary substrate.

For Operations teams at Corporate Banking firms, each hallucination has a direct operational consequence in the message template, correspondent-onboarding spec, or implementation guidance. The Panel's testing surfaces Fedwire hybrid postal address schema over-specification. Where these errors flow into a deliverable, the exposure is address-field rejects on Fedwire at go-live, remediation across origination systems, and re-run of bilateral UAT cycles. The pattern is uniform across the set: the AI returns a confident, sourced-looking answer that conflicts in a load-bearing specific with the regulator's verbatim text, and the error survives a first-pass review precisely because the surface form is plausible.

The Panel records each hallucination with the regulator's primary substrate held as the anchor, so the corrective text is available alongside the failure.

The Specialist Panel records the citation IDs as follows: RLB-H-INT-BIS-CPMI-ISO-20022-HARMONISATION-UPDATED-2026-Q010-Opus47 (Claude Opus 4.7 (web search on), Schema Over-Specification). Each citation links to the verbatim regulator-issued source text, the tested AI question, and the recorded AI response, so the Panel's assessment is traceable end to end. For operations teams at corporate banking firms, the citation IDs operate as a reference index: when an AI answer in the working draft matches a known Panel finding, the cited regulator text is already available as the corrective anchor.

The full per-finding analysis cards, including the audience-specific impact statement, sit on the cell's detail surface for sign-off use.

Technology & Data teams at Retail Banking firms implementing the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) are increasingly using AI to design message-schema validators, generate field-mapping logic for Fedwire and CHAPS rails, and draft API documentation for connected clients. The same tools build address-field parsing components for payment engines.

Two frontier AI models tested by the RLB Specialist Panel on the workflows retail-banking technology and data teams use to support advice on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) produced one discrete hallucination bound to regulator-issued source text. The Panel records a single recurring failure class: Schema Over-Specification across the set. Questions were prepared by the Specialist Panel based on real practical AI usage in the workflows retail-banking technology and data teams use AI for, and each finding is bound to verbatim regulator-issued source text held as primary substrate.

For Technology & Data teams at Retail Banking firms, each hallucination has a direct operational consequence in the message-schema validator, field-mapping component, or API specification. The Panel's testing surfaces Fedwire hybrid postal address schema over-specification. Where these errors flow into a deliverable, the exposure is non-compliant Fedwire messages, rejected transactions at clearing, and a rework cycle spanning data-mapping and payment-engine logic. The pattern is uniform across the set: the AI returns a confident, sourced-looking answer that conflicts in a load-bearing specific with the regulator's verbatim text, and the error survives a first-pass review precisely because the surface form is plausible.

The Panel records each hallucination with the regulator's primary substrate held as the anchor, so the corrective text is available alongside the failure.

The Specialist Panel records the citation IDs as follows: RLB-H-INT-BIS-CPMI-ISO-20022-HARMONISATION-UPDATED-2026-Q010-Opus47 (Claude Opus 4.7 (web search on), Schema Over-Specification). Each citation links to the verbatim regulator-issued source text, the tested AI question, and the recorded AI response, so the Panel's assessment is traceable end to end. For technology & data teams at retail banking firms, the citation IDs operate as a reference index: when an AI answer in the working draft matches a known Panel finding, the cited regulator text is already available as the corrective anchor.

The full per-finding analysis cards, including the audience-specific impact statement, sit on the cell's detail surface for sign-off use.

Technology & Data teams at Payment Institutions building cross-border payment infrastructure under the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) are increasingly using AI to design address parsers, build message-schema components for Fedwire-connected rails, and generate API documentation for connected clients. The same tools draft technical specifications for white-label cross-border payment products.

Two frontier AI models tested by the RLB Specialist Panel on the workflows payment-institution technology and data teams use to support advice on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) produced one discrete hallucination bound to regulator-issued source text. The Panel records a single recurring failure class: Schema Over-Specification across the set. Questions were prepared by the Specialist Panel based on real practical AI usage in the workflows payment-institution technology and data teams use AI for, and each finding is bound to verbatim regulator-issued source text held as primary substrate.

For Technology & Data teams at Payment Institutions, each hallucination has a direct operational consequence in the address parser, message-schema component, or API specification. The Panel's testing surfaces Fedwire hybrid postal address schema over-specification. Where these errors flow into a deliverable, the exposure is non-compliant Fedwire messages flowing through sponsor-bank governance, multi-team remediation across release cycles, and a contractual breach finding with the sponsor bank.

The pattern is uniform across the set: the AI returns a confident, sourced-looking answer that conflicts in a load-bearing specific with the regulator's verbatim text, and the error survives a first-pass review precisely because the surface form is plausible. The Panel records each hallucination with the regulator's primary substrate held as the anchor, so the corrective text is available alongside the failure.

The Specialist Panel records the citation IDs as follows: RLB-H-INT-BIS-CPMI-ISO-20022-HARMONISATION-UPDATED-2026-Q010-Opus47 (Claude Opus 4.7 (web search on), Schema Over-Specification). Each citation links to the verbatim regulator-issued source text, the tested AI question, and the recorded AI response, so the Panel's assessment is traceable end to end. For technology & data teams at payment institutions, the citation IDs operate as a reference index: when an AI answer in the working draft matches a known Panel finding, the cited regulator text is already available as the corrective anchor.

The full per-finding analysis cards, including the audience-specific impact statement, sit on the cell's detail surface for sign-off use.