Unauthorized group gains access to Anthropic's cybersecurity tool Mythos
Original: Unauthorized group has gained access to Anthropic’s exclusive cyber tool Mythos, report claims
Why This Matters
Highlights security vulnerabilities in AI model distribution and access controls
An unauthorized group has reportedly gained access to Anthropic's Mythos cybersecurity AI tool through a third-party vendor, according to Bloomberg. The group accessed the tool on its announcement day by guessing its online location based on Anthropic's model formats.
Bloomberg reports that members of a private online forum have gained unauthorized access to Anthropic's Mythos cybersecurity tool through a third-party vendor environment. The group, which includes someone employed at an Anthropic contractor, accessed the tool on the same day it was publicly announced by making an educated guess about its online location. Members are part of a Discord channel that seeks unreleased AI models and have been using Mythos regularly, providing screenshots and live demonstrations as evidence. Anthropic confirmed they are investigating the report but found no evidence their systems were impacted. Mythos was released to select vendors including Apple under Project Glasswing to prevent misuse by bad actors, as the tool could potentially be weaponized against corporate security.