<- Take me back to my Compliance x Corporate Banking (SG) overview
Executive Summary
Compliance teams at Singapore corporate banks operate MAS Notice 637 as a daily working instrument: compliance monitoring of capital adequacy ratios, ongoing assessment of capital instrument eligibility, regulatory change capture, and reporting to MAS on the Reporting Bank's adherence to the framework. Across the two findings in this cell, an AI model fabricated a parallel holding-company notice ('Notice FHC-N637') that does not exist, and misrepresented MAS's editorial convention for yellow highlighting in amendment PDFs.
For corporate banking compliance, both failures translate into compliance positions captured against the wrong regulatory basis: a fabricated parallel notice in a group-perimeter compliance assessment, and a wrong reading of editorial annotation as live regulatory text in the amendment change log.
How AI gets this regulation wrong
Both findings are inference drift: the AI committed to confident answers on questions where the regulator's own text resolves the issue. The first fabricates an instrument; the second misreads a convention. Both produce silent failures in the deliverable, with no signal to the compliance officer that the underlying claim has no basis in MAS's published text.
| AI's Failure Mode | Count | Affected findings |
|---|---|---|
| Exposed Fabrication | 2 | Finding#1 · Finding#2 |
What that means for your practice
For corporate banking compliance, both findings cluster on the same risk: regulatory enforcement exposure when MAS's own text resolves a question differently from the position captured in the bank's compliance records. A compliance assessment that documents the regulatory perimeter for a financial holding company by reference to a fabricated parallel notice, or a change log that captures highlighted annotation as live amendment text, both create compliance records that MAS could readily test against its own published instruments and find inconsistent.
| Risk Impact | Count | Affected findings |
|---|---|---|
| Regulatory enforcement / professional liability exposure | 2 | Finding#1 · Finding#2 |
When this affects Compliance teams at Corporate Banking firms
Corporate banking compliance teams apply Notice 637 across compliance monitoring of regulatory capital ratios, capital instrument eligibility memoranda, group-structure compliance assessments (particularly when the corporate bank sits under a Singapore financial holding company), regulatory change capture for Notice 637 amendments, and reporting on the bank's adherence under MAS supervisory engagement.
The two findings in this cell map onto the most common questions a corporate banking compliance team puts to an AI tool. First, what regulatory instrument governs capital adequacy at the financial holding company level (and whether it is a separate notice or a scope rule within Notice 637 itself). The AI's fabrication of a parallel notice is the kind of crisp, citable output that flows directly into a perimeter-compliance memo. Second, how to log items from a MAS tracked-change amendment that uses yellow highlighting.
The AI's wrong reading causes the change-capture process to record annotation text as a live regulatory obligation, embedding the error into the bank's regulatory change register.
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 corporate banking compliance teams should anticipate when AI tools are used for Singapore prudential research. The model fabricates instruments where the regulator has used scope rules within Notice 637 itself, and the model imposes a general convention onto an editorial annotation device that the regulator's own text explains.
For compliance records, this matters because both errors are silent in the output. A compliance memo on the financial holding company perimeter that cites the fabricated 'Notice FHC-N637' reads as a correctly sourced compliance position. A regulatory change register entry that captures highlighted annotation as a live amendment obligation reads as a routine change-capture entry. The compounding risk is that errors of this kind become part of the bank's compliance corporate record, where MAS inspection or internal audit review can test them against the source notice and find them inconsistent.
What your team should do
Corporate banking compliance teams should treat AI tools as a research-prompt generator on Notice 637 work, with a mandatory verification step against MAS's published text before AI output enters a compliance memo, regulatory change register, or supervisory submission. The findings in this cell concentrate on the two question types most exposed in compliance work: instrument identification and amendment-text reading.
Practical safeguards: (a) every MAS instrument citation entering the compliance record must be matched to the MAS publications portal listing. (b) When capturing changes from a MAS amendment with tracked-change conventions, pull the reading-convention paragraph of the amendment (paragraph 3 in the 2025 amendment) before logging items. (c) Build into the compliance team's AI-use guardrails a specific carve-out for instrument-identification and amendment-interpretation queries: these are precisely the question types where this testing shows the AI produces confident wrong answers.
(d) For group-perimeter assessments, anchor the position on Notice 637 paragraph 11.2.2 (the scope rule), not on AI-supplied references to parallel instruments.
Where AI tools support compliance work in this practice: outlining compliance memos, identifying which Parts of Notice 637 are likely relevant to a particular monitoring question, drafting first-pass change logs for verification against the actual amendment text, and surfacing cross-references between Notice 637 and adjacent MAS instruments for compliance review.
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.