On Anthropic’s Mythos Preview and Project Glasswing

The blog post discusses Anthropic’s Mythos Preview and Project Glasswing, warning that advanced AI models are increasingly capable of finding, chaining, and turning software vulnerabilities into operational exploits. While defenders currently retain some advantage—fixing vulnerabilities is easier than fully exploiting them—the author argues that as models improve and proliferate, zero-day discovery and automated exploit development will become far more common, raising urgent defensive and policy challenges.