Anthropic is set to let European banks access its cybersecurity tool Claude Mythos Preview, Reuters has reported.
Citing three people familiar with the matter, it reported that the AI company plans to expand access to European banks “soon”, letting them test software for vulnerabilities with the tool that has worried the financial services sector worldwide.
Mythos Preview was released earlier in April and caused concern among banks and governments almost immediately. At launch it had already found “thousands” of major vulnerabilities across software including all major operating systems and web browsers, Anthropic said.
Days later, US Treasury secretary Scott Bessent called a meeting with top bankers to discuss the threat it posed to the sector, followed closely by financial regulators in the UK and EU.
Anthropic has tightly controlled the software’s rollout to ensure it is not accessed by malicious actors; currently only certain US banks have access to the software, meaning European banks are having to plan solutions without knowing how affected they will be by its use.
One of the people speaking to Reuters said they could gain access within days, while another said it could be days or weeks.
Despite the controlled rollout, Mythos might not be as secure as advertised. Bloomberg News reported on Tuesday that a small number of unauthorised users accessed the AI platform on the day of its limited release in April, and have continued to use the model regularly since then. A company spokesperson said: “We’re investigating a report claiming unauthorised access to Claude Mythos Preview through one of our third-party vendor environments.”
