<- Take me back to my Company Secretaries (SG) overview
Executive Summary
Company secretaries at Singapore-licensed banks and their parent groups are the operating layer for board-level reporting on MAS Notice 637 compliance, group-structure governance, and amendment tracking. Across the two findings in this cell, an AI model with web search fabricated a parallel holding-company notice ('Notice FHC-N637') that does not exist, and misrepresented the meaning of yellow highlighting in MAS amendment PDFs.
For the company secretarial function, both failures land directly on board-paper preparation and on the integrity of governance records: a fabricated parallel notice cited in a board paper misleads the board on group regulatory architecture; a wrong reading of the amendment's editorial convention causes the secretarial function to log annotations as operative regulatory text in the board's compliance register.
How AI gets this regulation wrong
Both findings are inference drift: the AI committed to a specific answer where the regulator's own published text resolves the question. The first finding invents a regulatory instrument; the second mischaracterises a MAS editorial convention. Both failure modes are confident in tone and structurally plausible, which is exactly the failure profile most dangerous to a company secretarial workflow, where the output is being read for transcription into governance records rather than for substantive challenge.
| AI's Failure Mode | Count | Affected findings |
|---|---|---|
| Exposed Fabrication | 2 | Finding#1 · Finding#2 |
What that means for your practice
For company secretaries supporting Singapore-licensed banking groups, both findings translate into the same operational risk: governance records and board papers that capture the wrong regulatory position. A board paper that documents the regulatory perimeter for the group's financial holding company by reference to a fabricated parallel notice, or a compliance register that logs annotation text as a live regulatory obligation, both produce a record that the board has been advised on a position the regulator has not taken.
| Risk Impact | Count | Affected findings |
|---|---|---|
| Regulatory enforcement / professional liability exposure | 2 | Finding#1 · Finding#2 |
When this affects Company Secretaries
Company secretaries handle MAS Notice 637 across board paper drafting on capital adequacy compliance, regulatory change logs maintained for the audit committee, group-structure governance records (particularly when a Singapore banking group operates under a financial holding company), and minute-book records of board approvals tied to capital instruments issued under Notice 637's eligibility criteria. Each of these documents needs to record the correct regulatory instrument, the correct paragraph, and the correct amendment-status reading.
The two findings in this cell map onto the most routine questions a company secretarial team puts to an AI tool in this practice. First, what regulatory instrument applies to a financial holding company in the group structure (and whether it is a separate notice or a scope carve-out within Notice 637 itself). The AI's fabrication of a parallel notice is the kind of clean, citable answer that flows straight into a group-structure memorandum for the board.
Second, how to log items from a MAS amendment that uses yellow highlighting; here the AI's wrong characterisation produces a register that treats annotations as live obligations, embedding an error into the compliance change log that the secretarial team itself maintains.
The findings at a glance
The table below lists each finding from the AI testing on MAS Notice 637 in this cell, showing the topic, the AI's failure mode, and the citation identifier.
Aggregate impact
Considered together, the two findings describe a generation pattern that company secretaries should expect in AI-assisted work on Singapore banking notices. The model fabricates instruments where the regulator has instead used a scope-defined carve-out, and the model imposes a general convention onto an editorial annotation device that the regulator's own text explains.
For governance records and board reporting, this matters because both errors are silent in the deliverable. A board paper that cites a fabricated 'Notice FHC-N637' as the applicable instrument for the group's financial holding company does not flag the fabrication; it reads as a correctly sourced governance record. A change-log entry that treats yellow-highlighted text in the amendment as a live compliance obligation does not flag the convention error; it reads as a routine regulatory-change capture.
The compounding risk for the secretarial function is that errors of this kind become part of the corporate record, where they are subsequently referenced as authority for further governance decisions.
What your team should do
Company secretarial teams supporting Singapore-licensed banking groups should treat AI tools as draft-paper generators on Notice 637 topics, with a mandatory verification step against MAS's published text before any output enters a board paper, compliance register, or minute-book record. The findings in this cell show that the failure modes lie precisely in the question types most attractive to an AI-assisted workflow: instrument identification (where the AI invented a notice number) and amendment-convention reading (where the AI imposed a general drafting convention onto a MAS-specific annotation device).
Practical safeguards for the secretarial function: (a) every MAS instrument citation in a board paper or governance record must be matched against the MAS publications portal; an AI's output should be treated as a search prompt only. (b) When AI tools describe a feature of a MAS amendment PDF, pull the reading-convention paragraph of the amendment (paragraph 3 in the 2025 amendment) before logging the item in the compliance change register. (c) Maintain a short internal catalogue of confirmed MAS instruments applicable to the group; this gives the secretarial function a quick check against AI-suggested instrument names.
Where AI tools deliver value in this practice: drafting board-paper structures, capturing summaries of Notice 637 architecture for review against the source notice, and producing first drafts of regulatory change logs that the team subsequently verifies against the actual amendment text. The risk concentrates in the moment the AI's specific outputs (instrument numbers, convention statements) are transcribed into governance records without verification.
How RLB Can Help
RegLeg's published Hallucination Research is available as a free pre-flight check for Singapore banking practitioners working across MAS-supervised entities. Before relying on AI-assisted output for regulatory interpretation, compliance advice, or capital-instrument structuring, practitioners can consult the research to identify where AI tools have demonstrably mis-stated the rules: invented instruments, misread editorial conventions, outdated paragraphs presented as current. The research covers specific MAS instruments and surfaces the exact questions where AI tools have failed, making it a practical reference rather than a general caution.
For firms where multiple teams are working the same regulatory portfolio, RegLeg offers bespoke deep-dives into individual MAS instruments. These engagements go beyond the published findings to examine the full pattern of AI failure modes relevant to the instrument: the question types, the failure mechanisms, and the risk implications for compliance, risk, treasury, legal, and reporting work. The output is designed to be shared across functions and used as a durable reference, reducing duplicated due-diligence effort and creating a consistent internal standard for AI-assisted regulatory work.
RegLeg also develops training and CPD-aligned content for Singapore banking teams. The material translates the failure-mode catalogue into practical guidance on the classes of error practitioners should watch for: confabulated cross-references, version confusion between superseded and current instruments, jurisdiction bleed between superficially similar regimes, and inference-driven elaboration that overstates what an instrument actually requires. Separately, RegLeg offers a confidential review of a firm's existing AI-use policy against the failure-mode catalogue, identifying gaps between the policy's assumptions and the documented evidence of how AI tools perform on Singapore prudential questions in practice.