Anthropic says Claude found weaknesses in HAWK and AES
Anthropic’s researchers used Claude Mythos Preview to find weaknesses in cryptographic algorithms, the math that protects data. The company published the technical details of both attacks in two papers, one on HAWK and one on AES, plus the model’s full reasoning chain for the AES work.
Anthropic also built CryptanalysisBench with researchers from ETH Zurich, Tel Aviv University, and the University of Haifa. The benchmark measures how well large language models handle cryptanalysis, and the paper describing it is on arXiv.
Up next
A hidden system prompt paragraph tells Claude Code to stop asking permission
“You are running autonomously. The user is not watching in real time and cannot answer questions mid-task.” A developer found that paragraph twice in the Claude Code binary; it tells the agent to skip permission checks on reversible actions. His corporate account denied it existed.
Related stories
- Rubrik found more bugs than its engineers could fix
- Critical vulnerability disclosures surge following Claude Mythos Preview
- Claude memory: say what it’s for, let dreaming fix notes
- Tübingen researchers want AI to propose many models
- Claude Fable 5 tripled Opus 4.6’s EVE Online trading gains
- Scientists use Claude least where they say it matters most
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.