The Bank of England, the Financial Conduct Authority, and HM Treasury are all in the same room right now. Or will be, within the next two weeks. That alone should tell you how seriously Britain is taking Anthropic's Mythos.

According to the Financial Times, first reported Saturday, UK financial regulators have launched urgent discussions with the National Cyber Security Centre to assess the threats posed by Claude Mythos Preview. Representatives from every major British bank, insurer, and financial exchange are expected to attend a classified briefing within the fortnight.

Reuters confirmed the scope: this is not a routine consultation. This is the financial sector's equivalent of pulling the fire alarm.

Why Britain Is Panicking

Mythos Preview can find zero-day vulnerabilities in software systems at a speed and scale that no human security team can match. Anthropic has been so alarmed by its own model's capabilities that it refused to release it publicly, instead offering limited access only to major tech companies and government agencies through its Project Glasswing program.

The problem, as the FT makes clear, is that if Anthropic's model can find these vulnerabilities, it is only a matter of time before other models can too. And unlike Anthropic, the next lab might not choose responsibility over revenue.

Britain's financial sector runs on legacy software stacks that predate the smartphone era. Some of these systems were built in the 1990s and patched forward. A model that can autonomously discover exploitable flaws in that kind of infrastructure is not a theoretical risk. It is a specific, measurable, and imminent one.

The Mythos Domino Effect

Britain is now the fourth major jurisdiction to hold emergency-level discussions about Mythos. The US Treasury and Federal Reserve went first. Canada's central bank followed days later. Now London.

Notice the pattern: central banks and financial regulators, not tech regulators. That is significant. When the people who oversee the money supply start treating an AI model as a systemic risk, the conversation has moved well beyond "should we regulate AI" and into "how do we survive it."

What Happens Next

The briefing will likely result in mandatory vulnerability assessments for any UK financial institution running AI-adjacent infrastructure. Expect new guidance from the FCA on AI model risk within weeks, not months.

The bigger question is whether Anthropic's competitors are building similar capabilities in secret. OpenAI is reportedly developing its own cybersecurity-focused model. Google DeepMind has been conspicuously quiet. And the Chinese labs are not exactly going to announce what they can do.

Anthropic chose to disclose Mythos's capabilities and restrict access. For that, it deserves credit. But disclosure does not solve the underlying problem: AI models are now capable of finding and potentially exploiting weaknesses in the systems that move the world's money. And that genie is never going back in the bottle.