Bank of England, FCA, and NCSC convene emergency talks with major UK banks and insurers to assess systemic risks of Anthropic's Claude Mythos AI model

The Bank of England, the Financial Conduct Authority (FCA), and HM Treasury officials are holding urgent talks with the National Cyber Security Centre (NCSC) to assess potential vulnerabilities in critical financial IT systems exposed by Anthropic's latest AI model, Claude Mythos. Representatives from major British banks, insurers, and exchanges are expected to be briefed on the cybersecurity and systemic risk implications. The convening is understood to follow high-level concern — including at the US level, where Treasury Secretary Bessent and Federal Reserve Chair Powell separately warned bank CEOs about the model's capabilities — over Claude Mythos's unprecedented abilities in areas relevant to critical infrastructure attack vectors. Unlike previous AI risk assessments by UK financial regulators, which have focused primarily on conduct risks (such as model bias in lending decisions or AI-generated financial advice), this response is centred on systemic cyber risk: the concern that a sufficiently capable AI model could be used to identify and exploit vulnerabilities in the interconnected IT systems underpinning UK financial market infrastructure. The NCSC — the UK government's national authority on cybersecurity, operating under GCHQ — has a specific mandate to protect critical national infrastructure, and its involvement alongside the BoE and FCA signals that this is being treated as a potential systemic rather than firm-level risk. The urgency of the convening, within days of the model's release, reflects a new phase in how UK regulators approach frontier AI: from post-hoc guidance to pre-emptive crisis-style risk assessment.