On 7 April, Anthropic published research on Claude Mythos Preview, a general-purpose language model that the company says can autonomously discover and exploit software vulnerabilities. Rather than releasing it publicly, Anthropic is distributing Mythos through a restricted program called Project Glasswing to selected technology vendors for defensive research. Reactions have split between alarm ov...
The strongest version of this narrative is that AI-assisted vulnerability discovery is accelerating, and organizations must adapt or risk being overwhelmed. Anthropic’s Mythos Preview, while unproven by independent verification, serves as a harbinger of this shift. The claims—such as finding decades-old bugs in mature codebases—are plausible given the trajectory of AI capabilities, and the benchmarks, though self-reported, align with broader trends in AI-driven security research. The article rig...