/
RSS Feed
On April 17 the three federal banking agencies wrote generative and agentic AI out of model-risk guidance and promised a request for information; six months later that RFI is still unpublished and Reuters reports examiners asking banks the bluntest question in supervision: how do you turn your AI agents off? Two bears dig into the April carve-out, the four-question examiner checklist — limits, audit trails, human involvement, off-switches, foreign models — the Mythos backdrop, and the three-regulator split between US exam-room pressure, the ECB’s October 31 deadline, and the UK’s get-in-a-room approach.

Leave a Reply