Podcast Episode
During testing, Mythos found thousands of high-severity vulnerabilities in every major operating system and web browser, including some that had gone undetected for decades despite repeated human-led security audits. Logan Graham of Anthropic told PBS NewsHour that the model "excels in executing extensive tasks akin to those typically performed by a human security researcher over an entire day."
In the United States, Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bessent summoned major bank CEOs to discuss the model's risk profile. The Bank of Canada has also convened meetings with major lenders. IMF Managing Director Kristalina Georgieva warned that the risks "have been growing exponentially" and called for stronger safeguards to protect financial stability.
Anthropic Withholds Claude Mythos AI Model as UK Regulators Scramble to Assess Cyber Risks
April 14, 2026
0:00
2:48
Anthropic has unveiled its most powerful AI model yet, Claude Mythos Preview, but is refusing to release it publicly due to its extraordinary ability to find and exploit software vulnerabilities. UK financial regulators are now holding urgent talks to assess the risks, while US officials have summoned bank CEOs over the model's threat to financial stability.
A Model Too Powerful to Release
Anthropic announced Claude Mythos Preview on 7 April 2026 as part of Project Glasswing, a cybersecurity initiative bringing together more than 40 major technology companies to scan and secure critical software infrastructure. In an unprecedented move for the AI industry, the company declared the model would not be made publicly available, citing its extraordinary capability to autonomously discover and exploit software vulnerabilities.During testing, Mythos found thousands of high-severity vulnerabilities in every major operating system and web browser, including some that had gone undetected for decades despite repeated human-led security audits. Logan Graham of Anthropic told PBS NewsHour that the model "excels in executing extensive tasks akin to those typically performed by a human security researcher over an entire day."
Project Glasswing Partners
The initiative includes participation from Amazon Web Services, Apple, Google, Microsoft, Nvidia, JPMorgan Chase, Broadcom, Cisco, CrowdStrike, Palo Alto Networks, and the Linux Foundation. Anthropic has committed up to $100 million in usage credits and $4 million in direct donations to open-source security organisations. Partner organisations are using Mythos exclusively for defensive security work to identify code vulnerabilities before they can be exploited.Global Regulatory Response
The model's capabilities have triggered urgent regulatory action across multiple continents. In the UK, the Bank of England, Financial Conduct Authority, and UK Treasury are holding urgent talks with the National Cyber Security Centre and major financial firms. The UK's AI Security Institute published a formal evaluation finding that Mythos represents a meaningful "step up" in AI cyber capabilities.In the United States, Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bessent summoned major bank CEOs to discuss the model's risk profile. The Bank of Canada has also convened meetings with major lenders. IMF Managing Director Kristalina Georgieva warned that the risks "have been growing exponentially" and called for stronger safeguards to protect financial stability.
What Comes Next
Mythos sits above Anthropic's existing model lineup, with API pricing for gated partners set at $25 per million input tokens and $125 per million output tokens. No public release date has been announced, with speculation pointing towards either a near-term launch or integration into a broader Claude 5 release later in 2026.Published April 14, 2026 at 5:43pm
