- Anthropic will brief leading finance ministries and central banks on cyber vulnerabilities in the global financial system that Claude Mythos Preview has uncovered.
- The initiative traces to a request from Andrew Bailey, governor of the Bank of England and chair of the Financial Stability Board (FSB).
- Anthropic said last month that Mythos had found thousands of severe security flaws across all major operating systems and browsers.
- Only about 40 organisations currently have access to Mythos — including Amazon, Microsoft, and JPMorgan Chase. The White House recently asked Anthropic not to distribute the model more widely.
What Happened
Anthropic will brief leading finance ministries and central banks on cyber vulnerabilities in the global financial system uncovered by its Claude Mythos Preview model, The Decoder reported on Monday, citing the Financial Times. The initiative traces to a request from Andrew Bailey — governor of the Bank of England and chair of the Financial Stability Board (FSB), which brings together financial regulators from the G20 nations.
Why It Matters
The briefing is one of the most concrete operational examples to date of a frontier AI model being treated as a national-security and financial-stability asset rather than a consumer product. The pattern parallels how nuclear, biological, and certain cyber capabilities are governed: restricted distribution, government-coordinated assessment, controlled disclosure.
The International Monetary Fund warned recently that new AI models could turn cyber risks into a “macro-financial shock.” The FSB is preparing a report on AI use in the financial system, due out next month. The Anthropic briefing arrives ahead of that publication and is likely to inform its findings.
Technical Details
Anthropic said last month that Mythos had found thousands of severe security flaws across all major operating systems and browsers — a claim consistent with the Carnegie Mellon ExploitBench results published last week showing Mythos can autonomously develop V8 browser exploits, and the UK AISI evaluation showing Mythos is the first model to clear both AISI cyber ranges. About 40 organisations currently have access to Mythos, including Amazon, Microsoft, and JPMorgan Chase. The White House recently asked Anthropic not to distribute the model more widely.
Andrew Bailey’s dual role — as Bank of England governor and FSB chair — gives the request unusual operational weight. The FSB coordinates financial-stability policy across the G20 and includes the U.S. Treasury, Federal Reserve, European Central Bank, Bank of Japan, People’s Bank of China, and other major central banks.
Who’s Affected
The G20 financial-stability community — central banks, finance ministries, and major banking supervisors — gains coordinated visibility into a class of cyber risk most lack the technical capacity to assess independently. Anthropic gains an unusual policy-distribution moat: the model is operating as a regulatory-grade input, not a commercial product, at the FSB level. Other frontier AI labs — OpenAI, Google DeepMind, xAI, Mistral — face the question of whether to pursue similar briefing arrangements. Defensive cybersecurity firms — CrowdStrike, Palo Alto Networks, SentinelOne, Microsoft Defender — may see structural shifts in the threat-intelligence stack as agentic capability flows through national-security channels first.
What’s Next
The FSB’s report on AI use in the financial system is expected next month. Industry-watchers should expect related publications from the European Central Bank’s IT Risk Outlook and the U.S. Office of Financial Research. Mistral CEO Arthur Mensch’s warning last week to a French commission of inquiry — about the strategic dependency created by allowing US-built models to scan European military code bases — anticipates a similar dynamic at the EU level. The cumulative regulatory framework for frontier-AI cyber capability appears to be coming into focus through mid-2026.
