vulnerability detection

The National Security Agency is testing Anthropic's Mythos AI model to find security flaws in Microsoft products. UK government evaluation shows the model can autonomously execute 32-step corporate network attacks—and 99% of discovered vulnerabilities remain unpatched.