Compliance teams at Payment Institutions operating under the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) are increasingly using AI to draft regulatory horizon-scanning records on adoption progress, generate correspondent-network readiness assessments, and validate the postal-address mapping in the firm's ISO 20022 message structure. The same tools prepare supervisor-facing descriptions of ISO 20022 readiness.
Two frontier AI models tested by the RLB Specialist Panel on the workflows payment-institution compliance officers use to support advice on the CPMI Harmonised ISO 20022 Data Requirements (Updated Report) produced three discrete hallucinations bound to regulator-issued source text. The Panel records two distinct failure classes, Numeric Drift and Schema Over-Specification across the set. Questions were prepared by the Specialist Panel based on real practical AI usage in the workflows payment-institution compliance officers use AI for, and each finding is bound to verbatim regulator-issued source text held as primary substrate.
For Compliance teams at Payment Institutions, each hallucination has a direct operational consequence in the horizon-scanning record, network-readiness assessment, or supervisor-facing readiness description. The Panel's testing surfaces ISO 20022 adoption rate conflation (RTGS vs faster payments), Fedwire hybrid postal address schema over-specification, and ISO 20022 adoption rate conflation (RTGS vs faster payments). Where these errors flow into a deliverable, the exposure is skewed correspondent-network readiness picture, over-specified vendor due-diligence criteria, and a discoverable error in the firm's regulatory record.
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-Q006-Opus47 (Claude Opus 4.7 (web search on), Numeric Drift); RLB-H-INT-BIS-CPMI-ISO-20022-HARMONISATION-UPDATED-2026-Q010-Opus47 (Claude Opus 4.7 (web search on), Schema Over-Specification); RLB-H-INT-BIS-CPMI-ISO-20022-HARMONISATION-UPDATED-2026-Q006-Sonnet46 (Claude Sonnet 4.6 (web search on), Numeric Drift). 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 compliance teams at payment institutions, 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.
An AI assistant collapsed two materially distinct ISO 20022 adoption rates, more than three-quarters for faster payment systems and approaching half for RTGS systems, into a single invented 79% figure applied to both. The gap between approaching half (~50%) and 79% for RTGS is not noise: it determines whether a Payment Institution benchmarking its correspondent network against market norms concludes its RTGS counterparts are laggards or in line with peers.
A Compliance team that accepts the AI's figure may under-pressure correspondents on faster payment readiness while over-assuring the board on RTGS migration progress, producing a skewed readiness picture that sits in the firm's regulatory horizon-scanning record until a counterpart or regulator points out the discrepancy.
Payment institution compliance functions are the buyer of record for vendor implementations of the CPMI data model on Fedwire-connected rails, and the team that signs off on the firm's postal address mapping in the ISO 20022 message structure. The AI's substitution of a structured mandatory tier, adding Building Number, Post Code, and Country Sub-Division, for the FRB Services FAQ's actual specification of country code + town name + optional free-format lines introduces a stricter-than-required format into compliance assessments, vendor due diligence, and any supervisor-facing description of the firm's ISO 20022 readiness.
The over-specification is the harder failure mode to catch because the team's instinct is to treat a stricter interpretation as the safer interpretation, when in this case it is simply wrong.
Claude Sonnet 4.6 with web search exhibited the same blended 79% adoption figure as the Opus 4.7 test on the matching question, collapsing the regulator's separately-stated faster-payment and RTGS rates into a single composite. The structural exposure for this audience is identical to the Opus variant: an AI assistant collapsed two materially distinct ISO 20022 adoption rates, more than three-quarters for faster payment systems and approaching half for RTGS systems, into a single invented 79% figure applied to both.
The gap between approaching half (~50%) and 79% for RTGS is not noise: it determines whether a Payment Institution benchmarking its correspondent network against market norms concludes its RTGS counterparts are laggards or in line with peers. A Compliance team that accepts the AI's figure may under-pressure correspondents on faster payment readiness while over-assuring the board on RTGS migration progress, producing a skewed readiness picture that sits in the firm's regulatory horizon-scanning record until a counterpart or regulator points out the discrepancy.
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.