Anthropic’s Mythos Will Force a Cybersecurity Reckoning—Just Not the One You Think

Anthropic unveiled Claude Mythos Preview, saying it can autonomously find vulnerabilities and assemble working exploit chains.

The company is limiting access through Project Glasswing to a small set of organizations for security testing and evaluation.

Security experts are split: some see a real leap in AI-assisted exploitation, while others say it mainly lowers the bar for attacks already possible.

If the claims hold, defenders may need faster patching, stronger red teaming, and tighter software security practices.