UK extends frontier AI curbs beyond banking sector

The UK government has published cross-industry guidance formally extending frontier AI risk management obligations to all businesses, not just regulated financial institutions. Every enterprise deploying frontier models in production now carries affirmative risk management duties under UK regulatory expectations, a meaningful threshold shift from the prior posture of voluntary best practice. The Prudential Regulation Authority had previously targeted banks specifically, naming Anthropic Mythos and GPT-5.5 as sources of significant disruption risk. Thursday's broader guidance moves that warning upstream, applying the same logic to any firm running frontier models regardless of sector. Essentially: (UK government, PRA) are converting frontier AI governance from optional hygiene into an enforceable expectation before formal legislation arrives. - The guidance covers all sectors, not only PRA-regulated banking and insurance firms. - Anthropic Mythos and GPT-5.5 were explicitly named in the prior bank-specific alert as disruption risks, establishing a named-model precedent. - The transition from best practice to regulatory expectation is the operative legal threshold change for corporate compliance teams. For firms already running frontier models, the guidance creates a paper trail problem: the UK has now established that these risks were foreseeable, which will bear directly on any future enforcement action or litigation.