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 accountant advising a CCP on whether Basel CET1 equity qualifies toward its LNAFE buffer needs to state the correct qualifying condition: the KC3 carve-out turns on whether inclusion is 'relevant and appropriate to avoid duplicate capital requirements.' AI tools tested on this question replaced that condition with an invented KC4 liquidity overlay — and then, when challenged, denied the carve-out exists in KC3 at all. A practitioner who adopts either of those framings in a compliance opinion, or in advice on how a CCP should structure its LNAFE calculation, has misstated the rule.
When a regulator or a peer reviewer checks the KC3 text, the error is immediately apparent — and the practitioner, not the AI, owns the sign-off.
KC3 states a single, clean minimum: LNAFE equal to at least six months of current operating expenses. AI tools tested on this question reconstructed that floor as a 'greater of' dual-track test — adding a scenario-analysis leg drawn from KC2 — producing a more complex and more demanding standard than the rule actually sets. For an accountant drafting a Principle 15 compliance opinion or reviewing an FMI's internal LNAFE policy, adopting the AI's framing inflates the stated minimum and misrepresents the KC3 text to the client.
If that opinion is then used to assess regulatory compliance, or is disclosed in regulatory correspondence after the 2025 Level 3 assessment, the discrepancy between the practitioner's characterisation and the actual KC3 wording is an exposure point.