Anthropic's Mythos AI: Dangerous Release After 7 Weeks

Anthropic plans to widely release its Mythos-level AI models within weeks, despite previously deeming them too dangerous. Just seven weeks ago, the company restricted access to these models, which can autonomously find and exploit zero-day vulnerabilities across major operating systems and browsers. Anthropic now states it has made swift progress on safety safeguards. However, developers and cybersecurity experts remain deeply unsettled. The UK's AI Security Institute previously found the Mythos Preview model achieved a 73% success rate on expert-level cybersecurity challenges. It also completed a 32-step corporate network attack simulation in 3 out of 10 attempts. What's interesting is the timing. This reversal comes shortly after Anthropic reportedly closed a $65 billion funding round. This development could significantly impact cybersecurity and the future of AI safety.