Anthropic Briefs Global Finance Watchdog on Claude Mythos AI Model

Anthropic, a U.S. artificial-intelligence startup, has built Claude Mythos, a language model that can surface previously unknown flaws in information-technology systems. Because those flaws could be weaponised by hackers, the model has drawn scrutiny from regulators and financial institutions. In May 2026 the Financial Stability Board (FSB)—the global finance watchdog chaired by Bank of England governor Andrew Bailey—scheduled a briefing with Anthropic to assess Mythos’s potential impact on cyber-defence and on the stability of the worldwide financial system. The International Monetary Fund (IMF) has likewise warned that rapid AI developments raise systemic risk and called for coordinated oversight.

The Guardian

• Anthropic – Provides limited access to Mythos for a select group of technology firms and banks, including Apple and JPMorgan Chase.

The Guardian

• Financial Stability Board (FSB) – Monitors global financial-system stability; its steering committee includes senior officials from the United States, United Kingdom, Australia and China.

The Guardian

• UK AI Security Institute (AISI) – Independent body that evaluates advanced AI models; issued its latest appraisal of Mythos.

The Guardian

• International Monetary Fund (IMF) – Highlighted cross-border cyber risk in a recent blog post.

The Guardian

• Goldman Sachs (CEO David Solomon) and JPMorgan Chase (CEO Jamie Dimon) – Executives who publicly acknowledged the model’s capabilities.

The Guardian

• Financial Conduct Authority (FCA) chief executive Nikhil Rathi – Represented UK regulators in discussions about AI-related cyber risk.

The Guardian

AISI’s assessment noted a “notable capability jump” in the preview version of Mythos. The model successfully completed the “cooling tower” cybersecurity test—a previously unsolved challenge—in three of ten attempts, marking the first such success for any AI model evaluated by AISI. The institute also reported that the length of autonomous cyber tasks performed by frontier models has doubled within months, indicating rapid progress in AI-driven hacking capabilities.

The Guardian

If exploited, Mythos-identified vulnerabilities could affect the interconnected infrastructure of banks, payment networks and technology providers worldwide. The FSB’s engagement reflects concerns that cyber risk transcends national borders and that inconsistent oversight could weaken the stability of the global financial system. The IMF’s warning underscores the need for harmonised regulatory responses.

The Guardian

The FSB spokesperson affirmed that the board “welcomes engagement with Anthropic and other firms on emerging and frontier risks to global stability.” AISI emphasized its ongoing development of tougher hacking tests to track AI progress. The IMF’s blog post warned that “cyber risk does not respect borders” and called for coordinated oversight. FCA chief executive Nikhil Rathi said AI developments have been a “significant topic of conversation” at recent IMF meetings and highlighted new UK guidance urging firms to “double down on core cyber hygiene,” including legacy-system reviews, detection mechanisms, governance, recovery planning and insurance. Goldman Sachs CEO David Solomon described himself as “hyper-aware” of Mythos’s capabilities, while JPMorgan CEO Jamie Dimon noted that AI has made cyber defence “harder” but could ultimately aid protection.

The Guardian

Cybersecurity experts cautioned that Mythos represents an evolution rather than a revolution in cyber threats. They pointed out that most data breaches continue to stem from well-known issues such as weak authentication and unpatched software, suggesting that existing hygiene practices remain essential.

The Guardian

Sources differ on the extent of Mythos’s threat. AISI’s test results highlight a breakthrough, whereas other experts argue that the model’s impact may be limited compared with traditional vulnerabilities. No public technical details of Mythos have been released, leaving a gap in independent verification of its capabilities.

The Guardian

• “welcomes engagement with Anthropic and other firms on emerging and frontier risks to global stability” — FSB spokesperson

The Guardian

• “Frontier AI’s autonomous cyber and software capability is advancing quickly: the length of cyber tasks that frontier models can complete autonomously has doubled on the order of months, not years,” — AISI

The Guardian

• “Cyber risk does not respect borders. As AI capabilities spread across countries, inconsistent oversight could weaken a globally interconnected system,” — IMF

The Guardian

• “Anthropic has acted fairly responsibly in engaging with authorities” — Nikhil Rathi, FCA chief executive

The Guardian

• “core cyber hygiene” — Nikhil Rathi, FCA chief executive

The Guardian

The FSB plans to incorporate findings from Anthropic’s briefing into its upcoming policy recommendations on AI-related cyber risk. The UK Treasury and regulators will monitor compliance with the newly issued cyber-hygiene guidance, while AISI prepares additional stress tests for frontier AI models. Anthropic is expected to continue limited collaborations with selected financial and technology firms while maintaining its decision not to release Mythos publicly.

The Guardian