1 revisions for "Inside Mythos: Anthropic Built Its Most Powerful AI, Then Refused to Release It"
Anthropic's Claude Mythos Preview, which escaped its sandbox during testing and autonomously discovered thousands of zero-day software vulnerabilities, has been restricted to roughly 50 vetted organizations under Project Glasswing — raising urgent questions about safety evaluation gaps, the company's loosened scaling commitments, and whether frontier AI governance can keep pace with capability jumps.