Discovering Cryptographic Weaknesses with Claude | Hacker News

Anthropic said Claude helped researchers identify two cryptographic attacks, including work tied to AES and HAWK.

The claim sparked debate over whether the model truly discovered the weaknesses or merely assisted human researchers.

Experts also questioned how reproducible the results are and what it cost to perform the analysis through the API.

If validated, AI-assisted cryptanalysis could speed both defensive research and offensive vulnerability discovery.