Briefings Blog

Compliance teams at Singapore investment-banking divisions are increasingly using AI to update the investment-banking regulatory-perimeter map for MAS Notice 637, draft deal-clearance regulatory-impact memoranda referencing the FHC framework, generate compliance-training summaries on the 31 December 2025 amendment, and prepare supervisor-facing position papers on group-capital obligations. 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 Compliance teams at Singapore investment-banking divisions the operational consequence is concrete. An IB deal-clearance memo that cites a fabricated MAS instrument would enter transaction documentation and surface on external diligence. A regulator-facing position paper that treats amendment annotation as substantive Notice text would mischaracterise the rule estate to MAS. Both errors are direct enforcement-risk events tied to AI output that was not bound to the regulator's 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.

Risk functions at Singapore retail-banking divisions are increasingly using AI to update the retail-banking regulatory-capital framework map, draft RWA classification notes for the 31 December 2025 amendment, generate stress-test documentation for consumer portfolios under MAS Notice 637, and prepare board-level dashboards on FHC-level capital obligations. 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 Risk functions at Singapore retail-banking divisions the operational consequence is concrete. A retail-banking regulatory-framework map that names a fabricated MAS notice would drive RWA classification and capital-buffer calibration through an instrument that does not exist. Risk-model documentation that captures amendment annotation as substantive Notice text would generate a versioning artefact that fails reconciliation against the published Notice. Both errors are direct model-governance issues.

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 Singapore retail-banking divisions are increasingly using AI to draft legal opinions on MAS Notice 637 amendment effects for product counsel, prepare consumer-disclosure language tied to capital-adequacy positioning, generate first-pass risk-of-non-compliance memoranda, and validate group-structure references in product documentation. 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 Legal teams at Singapore retail-banking divisions the operational consequence is concrete. A legal opinion that routes through a fabricated MAS instrument would not survive external counsel review. A consumer disclosure that builds on amendment annotation as substantive Notice text would misstate the rule estate to retail customers. Both errors expose the institution to written-record and conduct risk tied to AI output that was not bound to the regulator's 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.

Retail-banking compliance teams at Singapore-incorporated banks are increasingly using AI to update the retail-banking regulatory-perimeter map for MAS Notice 637, generate compliance-training summaries on the 31 December 2025 amendment, draft supervisor-facing letters on FHC-level capital obligations, and prepare the policy-register entry for the consolidated 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 Retail-banking compliance teams at Singapore-incorporated banks the operational consequence is concrete. A retail-banking compliance register that names a fabricated MAS notice would propagate the error into the bank's policy estate and into reporting to senior management. A training summary that treats amendment annotation as substantive Notice text would teach staff that rules apply when the regulator's cover note states they will not appear in the published Notice. Both errors are visible to MAS on review.

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.

Treasury teams at Singapore corporate-banking divisions are increasingly using AI to draft capital-instrument issuance memoranda against MAS Notice 637, generate ALM working notes on the amendment effects, prepare due-diligence packages for senior and subordinated debt issuance, and update group-capital reporting templates for the consolidated 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 Treasury teams at Singapore corporate-banking divisions the operational consequence is concrete. Internal capital-structure memoranda that name a fabricated MAS notice would surface in due-diligence packages and ratings-agency dossiers, and would not resolve to any MAS register entry. Template updates that capture amendment annotation as substantive Notice text would generate reconciliation issues against the published Notice. Both errors trace to AI output that was not verified against the regulator's 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.