Finance ministers, central bankers, and top financial executives are raising urgent concerns about a powerful new artificial intelligence model they fear could expose critical vulnerabilities in global banking systems.
At crisis meetings during this week's International Monetary Fund gathering in Washington DC, officials discussed the Claude Mythos model developed by Anthropic, which has demonstrated unprecedented capability in identifying cybersecurity weaknesses in major operating systems.
Canadian Finance Minister François-Philippe Champagne told reporters the situation warrants immediate attention from all finance ministers worldwide. "The issue we're facing with Anthropic is that it's the unknown, unknown," he said, contrasting it with more predictable geopolitical risks like the Strait of Hormuz. "This is requiring a lot of attention so that we have safeguards and processes in place to ensure the resiliency of our financial systems."
"It's serious enough that people have to worry. We have to understand it better, and we have to understand the vulnerabilities that are being exposed and fix them quickly." — CS Venkatakrishnan, Chief Executive of Barclays
Mythos represents Anthropic's latest advancement in its Claude AI system, which competes with OpenAI's ChatGPT and Google's Gemini. Earlier this month, developers testing AI models for "misaligned" tasks—those that could work against human values—described Mythos as "strikingly capable at computer security tasks."
Due to concerns that the model could surface old software bugs or find new ways to exploit system vulnerabilities, Anthropic has withheld Mythos from public release. Instead, the company has made it available through Project Glasswing to select technology giants including Amazon Web Services, Microsoft, and Nvidia, describing the initiative as "an effort to secure the world's most critical software."
On Thursday, Anthropic released a new version of its existing Claude Opus model, which will allow testing of Mythos' cybersecurity capabilities in less powerful systems.
While concerns about Mythos exceed previous AI model discussions, some cybersecurity experts question whether the alarm is justified, particularly since the model hasn't undergone widespread industry testing. The UK's AI Security Institute, which has published the only independent report on Mythos' capabilities, found it to be a powerful tool for identifying security holes in undefended environments but noted it wasn't dramatically better than Claude's predecessor, Opus 4.
Bank of England Governor Andrew Bailey emphasized the seriousness of the development: "We are having to look very carefully now what this latest AI development could mean for the risk of cyber crime. The consequence could be that there is a development of AI, of modeling, which makes it easier to detect existing vulnerabilities in core IT systems, and then obviously cyber criminals—the bad actors—could seek to exploit them."
The US Treasury has already raised the issue with major banks, encouraging them to test their systems before any potential public release of Mythos. Financial industry sources indicate another prominent US AI company may soon release a similarly powerful model without equivalent safeguards.
James Wise, partner at Balderton Capital and chair of the Sovereign AI unit—a venture capital fund backed by £500 million in government funding to invest in British AI companies—described Mythos as "the first of what will be many more powerful models" that can expose system vulnerabilities. "We hope the models that expose vulnerabilities are also the models which will fix them," he added.
As financial institutions prepare for what Barclays' CEO calls "the new world" of interconnected financial systems with both opportunities and vulnerabilities, governments and banks are being offered early access to Mythos to help protect their systems before any broader release.