The Governor of the Bank of England, Andrew Bailey, who chairs the global financial risk monitoring committee, has invited Anthropic to present to the finance ministries and central banks of the G20. The model that worried him is now being explained.
See also: Mythos proves powerful in vulnerability discovery but not elsewhere
Anthropic is preparing to brief the Financial Stability Board on the cyber vulnerabilities that the Mythos model has identified in the global financial system, the Financial Times reports. The briefing has been requested by Andrew Bailey and will be delivered to G20 finance ministers and central banks under the auspices of the board.
Bailey is the right person to make this move. In an April 15 speech at Columbia University, he cited Mythos along with the escalation in the Gulf as one of two events that have pushed cyber risks up the regulators’ rankings faster than any other category in recent years.
Mythos was announced last month but has not yet been released. Anthropic describes it as a cybersecurity model designed to expose long-standing vulnerabilities in browsers, infrastructure and software, claiming to have already found thousands of serious flaws in major operating systems and browsers.
When asked to develop working exploits of these vulnerabilities in internal testing, the model reportedly succeeded on the first try in over 83% of cases. The dual-use implications were clear to regulators before the model was presented.
See also: Japanese banks gain access to Anthropic's Mythos
The FSB briefing closes a regulatory cycle that began with Bailey’s speech and has since involved jurisdictions on three continents. UK banks received their own briefing on Mythos shortly after the remarks at Columbia. The Federal Reserve and the US Treasury Department convened the CEOs of major US banks about the same risk shortly afterwards.
Australia's securities regulator was added to the watchlist in early May. Eurozone finance ministers set their own access requirements, and Mythos has since been handed over to Japanese megabanks.
What the update won't solve is the issue of access. Mythos is currently available under "Project Glasswing," a controlled-access program created by Anthropic to limit who can run the model and on what.
About 40 to 50 organizations have early access, including AWS, Apple, Google, Microsoft, Nvidia, Cisco, and JPMorgan. Banking supervisors outside that list are pushing for either direct access or a regulatory-mediated equivalent, and the FSB meeting will be the first time these requests have been coordinated rather than handled nationally.
The political background is more complex than the technical. The model being reported to G20 financial regulators is a US-based AI system, the distribution and military access to which has been the subject of an ongoing dispute between Anthropic and the Trump administration.
See also: OpenAI Daybreak: The Answer to the Claude Mythos for Cybersecurity
Regulators will be aware that the company appearing before them has, in parallel, negotiated its export profile with Washington. Anthropic and the FSB had not responded to requests for comment by Monday afternoon, and the timing of the actual briefing has not been made public.
🔒 Protect your privacy with Proton VPN
Swiss VPN from the creators of Proton Mail — strict no-logs policy, strong encryption, and built-in NetShield that blocks ads, trackers, & malware.
- ✔ No-logs, based in Switzerland (except 14-Eyes)
- ✔ NetShield: blocks ads, trackers & malicious domains
- ✔ Covers all devices — free version available
The link is an affiliate link — SecNews may receive a commission at no additional cost to you. It does not affect the independence of our article writing.
