Compliance teams inside statutory boards and agencies (national payments councils, designated FMI oversight units, payments-policy desks at financial-services authorities) touch CPMI material at three points: the peer-jurisdiction tracker on regulator-led pilots, the inter-agency policy briefing on d224 alignment, and the response file for parliamentary or ministerial questions on payments. Two AI failures on this regulation hit those exact deliverables. Sonnet 4.6 denied the SARB pre-validation partnership documented in CPMI Brief No. 9, and Opus 4.7 returned a per-recommendation stakeholder taxonomy reconstructed from category labels rather than the d224 recommendation text.
A compliance team that lifts either AI output into a policy or ministerial deliverable is denying a documented regulator workstream or mis-scoping domestic regulatory action.
What the AI got wrong, and why it matters here
Both failures land on artefacts that are externally readable by counterpart agencies and by the political-level audience. Mistakes here propagate through the inter-agency channel before the body's own second line catches them.
Finding 1: SARB pre-validation partnership denied
Sonnet 4.6 denied any public CPMI statement names SARB as the pre-validation pilot partner. CPMI Brief No. 9 (November 2025) names SARB outright. Quoted into a peer-jurisdiction tracker or a parliamentary question response, the denial misrepresents a documented regulator workstream that the body's SA counterpart can confirm in print.
Citation: RLB-H-INT-BIS-CPMI-API-HARMONISATION-CROSS-BORDER-2024-Q007-Sonnet46.
Finding 2: Invented per-recommendation stakeholder taxonomy
Opus 4.7 returned a clean stakeholder taxonomy across d224's 10 recommendations, built from category labels rather than the recommendation text. A statutory-board policy memo written off that taxonomy mis-scopes domestic action: which recommendations require payments-act amendment, which require an oversight-perimeter rewrite, which fall to standards bodies.
Citation: RLB-H-INT-BIS-CPMI-API-HARMONISATION-CROSS-BORDER-2024-Q008-Opus47.
When this hits the policy calendar
Statutory-board compliance pulls CPMI material on three artefacts: the peer-jurisdiction pilot tracker, the inter-agency policy briefing on d224 alignment, and the parliamentary or ministerial question response file.
| Standing artefact | Where the AI risk surfaces | Failure mode |
|---|---|---|
| Peer-jurisdiction pilot tracker | Pilot-partner naming | Finding 1 |
| Inter-agency policy briefing on d224 alignment | Stakeholder-obligation mapping | Finding 2 |
| Parliamentary or ministerial question response | Both | Both |
Aggregate impact on the team
Both failures hit politically-sensitive artefacts. A factual error in a peer-jurisdiction tracker or a minister's brief carries reputational and inter-agency-trust cost that the body's own second line cannot reverse easily.
| Risk Impact | Count | Affected findings |
|---|---|---|
| 0 | ||
| 0 |
What this team should do
Treat AI output naming a CPMI pilot partner or mapping d224 obligations to domestic regulation as draft material requiring verification against the relevant CPMI brief by number and the d224 recommendation text before it enters a policy or ministerial deliverable.
Detection patterns to add to AI-review
- Pilot-partner statements must trace to a numbered CPMI brief.
- Obligation mapping on d224 must be verified against the recommendation text.
How RLB can help
RLB tracks AI failures on d224 and the CPMI brief series and refreshes the catalogue against live AI subjects on rotation. Statutory-board compliance teams can wire the catalogue into the policy-deliverable review step so these two failure shapes are caught before they reach the minister or a counterpart agency.