Operations teams at Retail Banking firms running cross-border payments programmes under the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) are increasingly using AI to build internal business cases for harmonisation investment, generate operational metrics for COO challenge sessions, and configure Fedwire payment-message templates. The same tools draft vendor due-diligence questionnaires on cross-border payment readiness.
Two frontier AI models tested by the RLB Specialist Panel on the workflows retail-banking operations teams use to support advice on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) produced two discrete hallucinations bound to regulator-issued source text. The Panel records two distinct failure classes, False-Negative Retrieval and Schema Over-Specification across the set. Questions were prepared by the Specialist Panel based on real practical AI usage in the workflows retail-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 Retail Banking firms, each hallucination has a direct operational consequence in the operations business case, COO challenge pack, or payment-message specification. The Panel's testing surfaces missing inquiry-rate and resolution-time benchmarks, and Fedwire hybrid postal address schema over-specification. Where these errors flow into a deliverable, the exposure is live STP failures, manual repair queues, and a re-run of UAT cycles mid-programme.
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-Q007-Sonnet46 (Claude Sonnet 4.6 (web search on), False-Negative Retrieval); 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 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.
This is the consolidated view of findings. Click the Citation IDs or 'see details →' on any item for the full details for each finding.
When an operations team builds the internal business case for ISO 20022 harmonisation investment, the Panetta speech figures, 1-3% of cross-border payments generating inquiries, 5-10 manual touchpoints per inquiry, up to 80% resolution-time reduction, are the kind of official benchmarks a CFO or COO challenge team will ask to see sourced. An AI tool that fails to surface these figures, or misattributes the 80% improvement to SWIFT or commercial banks rather than an official CPMI statement, leaves the team either citing the wrong source or unable to cite any source at all.
A business case built on unsourced or misattributed efficiency claims is vulnerable to challenge during investment approval and creates a credibility problem for the operations function sponsoring the programme.
An operations team configuring Fedwire payment messages under the hybrid postal address model needs to correctly implement the optional component as free-format lines of up to 70 characters each, not as structured fields (Street Name, Building Number, Post Code) drawn from generic ISO 20022 address conventions. An AI tool that inverts this requirement will produce a technical specification that looks credible but will cause STP failures or message repair queues at the receiving end when the optional component is populated with free-format content the spec has not accounted for.
In a multi-correspondent environment, a shared specification built on this error propagates across the full cross-border payments book before the first test message surfaces the problem.
Every finding on this page compares an AI subject's account of the rule against the regulator's verbatim text from the regulator's own portal. Both are linked. Each delta, its root causes, and impact analysis are documented and published with immutable Citation IDs.