Anthropic’s Mythos breach was humiliating
Anthropic's AI model Claude Mythos was accessed by unauthorized users despite the company's claims that it was too dangerous for public release. The breach occurred due to an 'educated guess' about the model's online location using information exposed in a separate breach.
Anthropic's rollout of Claude Mythos has been compromised, with a 'small group of unauthorized users' gaining access to the model. The breach happened when the group made an 'educated guess' about the model's online location using information from a Mercor breach. Anthropic is investigating, but the incident is embarrassing for a company that touts its AI safety and cybersecurity prowess. Security researchers say the breach was 'entirely imaginable' and that Anthropic should have anticipated it. The company had the means to spot the breach but apparently wasn't monitoring closely enough. The unauthorized users didn't use Mythos for cybersecurity tasks, which Anthropic claims is a 'lucky break' given the model's supposed capabilities.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.