Risk teams in corporate-bank cross-border payments scope concentration risk, settlement risk and FMI-connection risk against a narrow set of CPMI-published numbers: the count of operational fast payment systems, the live cross-border linkages, and the central-bank-versus-private operator mix. Two AI failures on this regulation hit that same number set in opposite directions: Opus 4.7 returns a 57-FPS 2025-monitoring-survey number with no operator-type breakdown, and Sonnet 4.6 returns the 70-plus headline correctly but denies that the operator-mix percentages exist.
The Tara Rice November 2023 speech (sp231115) is the primary source for the full set, with 70-plus operational, 14 cross-border-enabled, 24 in the five-year planning pipeline, 40% central-bank-operated and 35% privately operated. A second-line risk memo built off either AI answer enters the risk committee with the wrong universe size, no operator-mix differentiator and a forward-pipeline signal stripped out.
What the AI got wrong, and why it matters here
Both failures share the same operational pattern: a high-confidence answer that strips a signal load-bearing for risk differentiation. Neither is caught by a desk reviewer who is not already holding the sp231115 numbers in mind.
Finding 1: FPS universe compressed and operator mix erased
Opus 4.7 cited the 2025 monitoring survey at 57 (56 in one graph) operational fast payment systems and gave no operator-type breakdown. The Tara Rice November 2023 speech (sp231115) gives 70-plus operational, 14 cross-border-enabled, 24 in the five-year planning pipeline, 40% central-bank-operated and 35% privately operated. Risk memos calibrated against the AI answer enter committee with an undersized universe, no operator-mix differentiation, and the planning-pipeline forward indicator dropped from the input set.
Citation: RLB-H-INT-BIS-CPMI-API-HARMONISATION-CROSS-BORDER-2024-Q010-Opus47.
Finding 2: Operator-mix denied
Sonnet 4.6 cited the 70-plus headline correctly but denied that a precise central-bank-versus-private operator percentage is enumerated in the public Brief 10 summary. The Tara Rice November 2023 speech (sp231115) names 40% central-bank-operated and 35% privately operated. Removing the mix from the risk-memo input set collapses the central-bank-versus-private differentiation, which is the load-bearing signal for FMI-concentration analysis on cross-border rails.
Citation: RLB-H-INT-BIS-CPMI-API-HARMONISATION-CROSS-BORDER-2024-Q010-Sonnet46.
When this hits the risk calendar
Risk pulls CPMI material on three recurring deliverables: the FPS connectivity exposure memo, the cross-border payments FMI-concentration view, and the annual risk-appetite calibration for cross-border-payments business lines.
| Standing item | Where the AI risk surfaces | Failure mode |
|---|---|---|
| FPS connectivity exposure memo | Global FPS count, cross-border linkage count | Finding 1 |
| Cross-border-payments FMI-concentration view | Operator-mix differentiator | Findings 1 and 2 |
| Annual risk-appetite calibration | Planning-pipeline forward indicator and operator mix | Findings 1 and 2 |
Aggregate impact on the team
The same two failures collapse the operator-mix differentiation and the planning-pipeline forward signal, removing two of the three inputs risk-appetite calibration relies on.
| Risk Impact | Count | Affected findings |
|---|---|---|
| 0 |
What this team should do
Tag the FPS count and the operator mix as known-failure outputs. Any AI draft naming those numbers must be sent through a primary-source check against sp231115 and the latest CPMI cross-border monitoring brief before it lands in a second-line risk memo or a risk-appetite paper.
Detection patterns to add to AI-review
- Any FPS count must be tied back to Tara Rice sp231115 or to the most recent CPMI cross-border monitoring brief by number.
- Any denial that an operator-mix percentage exists must be cross-checked against sp231115 directly.
How RLB can help
RLB tracks the failure pattern on these specific numerical anchors across d224, the CPMI cross-border-payments brief series and the Tara Rice November 2023 speech, refreshed against live AI subjects on rotation. Risk teams in corporate banking can wire the catalogue into the AI-draft review step so the operator-mix line and the universe size never reach risk committee without a primary-source check.