Anthropic’s Mythos breach was humiliating

Anthropic, a company that emphasizes AI safety and the cybersecurity potential of its models, recently suffered a humiliating security breach. Despite claiming its Mythos model was too dangerous for public release, unauthorized users accessed it using simple methods involving insider knowledge and educated guesses. Experts suggest that Anthropic failed to anticipate foreseeable risks, despite having the capability to monitor and prevent such unauthorized access. This incident undermines the company's carefully curated image as a responsible leader in AI safety.

(Source：The Verge)