Anthropic is updating the Financial Stability Board on the findings of Mythos.

      Andrew Bailey, the Governor of the Bank of England and chair of the global financial-risk watchdog, has invited Anthropic to present to the finance ministries and central banks of the G20. The model that concerned him is the one currently being detailed.

      As reported by the Financial Times on Monday, Anthropic is set to inform the Financial Stability Board (FSB) about the cybersecurity vulnerabilities its Mythos model has discovered within the global financial system, according to sources familiar with the situation. This briefing was requested by Bailey and will be given to G20 finance ministries and central banks under the FSB's purview.

      Bailey is the ideal figure to make this request. In an April 15 address at Columbia University, he specifically mentioned Mythos, alongside the increased tensions in the Gulf, as two events that have rapidly elevated cybersecurity in regulators' risk assessments ‘more than any other category in recent years.’

      "It would be reasonable to think that the events in the Gulf are the most recent challenge to us in this world," Bailey stated, "until, I believe it was last Friday, you wake up to find that Anthropic may have discovered a way to fundamentally change the entire landscape of cyber risk."

      Announced last month but not yet released, Mythos is characterized by Anthropic as a cybersecurity model intended to reveal longstanding vulnerabilities in browsers, infrastructure, and software. The company claims it has already identified thousands of vulnerabilities classified as high-severity across major operating systems and browsers. During internal testing, the model reportedly succeeded in developing working exploits against these vulnerabilities over 83% of the time on its first attempt. The dual-use implications raised concerns for regulators even before the model became publicly available.

      The FSB briefing marks the culmination of a regulatory discussion that began with Bailey's speech and has since involved jurisdictions across three continents. Shortly after his remarks at Columbia, UK banks received their own Mythos briefing. The Federal Reserve and the US Treasury subsequently brought together the CEOs of major American banks regarding this risk. In early May, Australia’s securities regulator joined the list of interested parties. Euro-area finance ministers also expressed their own demands for access, and Mythos has since been provided to Japanese megabanks, as reported last week.

      However, the briefing will not address the access issue. Mythos is currently being distributed under "Project Glasswing," a controlled-access program established by Anthropic to restrict who can operate the model and for what purpose. Approximately 40 to 50 organizations have been granted early access, including AWS, Apple, Google, Microsoft, Nvidia, Cisco, and JPMorgan. Bank regulators outside this group have publicly advocated for either direct access or a mediated equivalent from regulators, and the FSB session will be the first opportunity for these requests to be coordinated rather than made individually.

      The political implications are more complex than the technical ones. The AI model being presented to G20 financial regulators is based in the US, and its distribution and military access have prompted ongoing disputes between Anthropic and the Trump administration. Regulators present will recognize that the company before them has been concurrently negotiating its export profile with Washington. By Monday afternoon, Anthropic and the FSB had yet to respond to requests for comment from Reuters, and the specific timing of the briefing has not been publicly confirmed.

      Mythos is the first AI system disclosed to the public that, according to its developer, has found exploitable vulnerabilities in “every major operating system and web browser.” The FSB briefing represents the initial occasion for the global financial supervision community to gather and discuss the practical ramifications of this discovery. The core question posed by Bailey during his Columbia speech regarding how much harder the model has made offensive actions in contrast to defensive measures remains a central issue.