Anthropic Limits Access to Powerful New Cybersecurity AI Model “Mythos”
Anthropic has restricted access to its latest model, Mythos, due to its unprecedented ability to autonomously identify and exploit hidden software vulnerabilities at an 80% success rate. Officials and cybersecurity experts are expressing concern over the dual-use risks of such advanced offensive capabilities, even as the company positions Mythos as a defensive tool for protecting critical infrastructure.
