Finance Ministers, Bankers Express Cyber Concerns Over Anthropic Mythos AI Model
A powerful new AI model, Anthropic's Claude Mythos, has become a central point of contention among global finance ministers, central bankers, and financiers, who fear its capabilities could undermine the security of international financial systems. Developers have stated the model is "strikingly capable at computer security tasks", revealing vulnerabilities in numerous major operating systems.
Canadian Finance Minister François-Philippe Champagne confirmed extensive discussions about Mythos at the recent International Monetary Fund (IMF) meeting in Washington DC. Champagne stated the issue was "serious enough to warrant the attention of all the finance ministers," highlighting the "unknown, unknown" nature of the threat compared to more geographically defined risks. He stressed the need for safeguards to ensure the resilience of financial systems.
Anthropic, which developed Mythos as part of its Claude AI system, has not publicly released the model. Instead, it has been provided to tech giants like Amazon Web Services and Microsoft through an initiative called Project Glasswing, described as an effort to "secure the world's most critical software."
While the UK's AI Security Institute, granted preview access, reported that Mythos could exploit systems with weak security, its independent assessment suggested the model was not dramatically superior to its predecessor, Claude Opus 4. However, the Institute's researchers acknowledged it was a powerful tool capable of finding numerous security flaws.
Barclays CEO CS Venkatakrishnan urged a deeper understanding of the vulnerabilities exposed by Mythos, calling for swift remediation. Bank of England Governor Andrew Bailey similarly emphasised the need to assess what this AI development could mean for cybercrime, particularly regarding cybercriminals exploiting system weaknesses identified by such models.
The US Treasury has encouraged major banks to test their systems in anticipation of any public release. This aligns with broader concerns that a similarly potent AI model could soon emerge from another prominent US AI company, potentially without comparable safeguards.
James Wise, chair of the UK's Sovereign AI unit, reiterated that Mythos is likely the first of many such powerful models capable of exposing system vulnerabilities, noting his unit's investment in British AI companies focusing on security and safety solutions, with the hope that models exposing flaws can also be leveraged to fix them.
