XBOW tests Anthropic's Mythos Preview for offensive security

XBOW tested Anthropic's Mythos Preview model and found it substantially better than prior models at finding vulnerability candidates, especially with source code. The model communicates with technical precision and shows strong promise in native-code analysis and reverse engineering. XBOW assembled a 10-expert team and used its internal benchmarking system to evaluate the model.