Anthropic’s Claude Mythos AI Sparks Concern Among G7 Finance Ministers This Week
Anthropic, a leading AI firm, has stated its new Claude Mythos model demonstrates superior capabilities in hacking and cybersecurity, prompting discussions among G7 finance ministers and central bankers, including at a recent International Monetary Fund (IMF) meeting in Washington DC.
The company claims Mythos can locate and exploit dormant bugs, even in decades-old code. Canadian Finance Minister François-Philippe Champagne described the technology as an "unknown unknown" warranting the attention of finance ministers, while Bank of England Governor Andrew Bailey noted the need to "look very carefully now what this latest AI development could mean for the risk of cyber crime." The EU is also reportedly engaging with Anthropic over its concerns.
Project Glasswing, an initiative by Anthropic, has granted 12 tech giants, including Amazon Web Services, Apple, Microsoft, Google, Nvidia, and Broadcom, access to Mythos. Anthropic stated that Mythos Preview has identified "thousands of high-severity vulnerabilities," some present for up to 27 years.
However, independent cybersecurity analysts and experts have not yet had widespread access for testing, leading to scepticism regarding the model’s actual performance. The UK’s AI Safety Institute concluded that while powerful, Mythos’s primary threat targets poorly defended systems. Its researchers stated, "We cannot say for sure whether Mythos Preview would be able to attack well-defended systems." Ciaran Martin, former head of the UK’s National Cyber Security Centre, noted that the claims have "really shaken people" but also highlighted the opportunity to use such tools to address underlying internet vulnerabilities.
Critics observe that capitalising on a mix of fear and excitement has become a consistent marketing strategy within the AI sector. The true extent of Mythos's capabilities remains to be fully verified by independent review, distinguishing verified claims from industry hype.
