Cryptographer Matthew Green discusses Anthropic's new cryptanalysis results, which include attacks on the signature scheme HAWK and reduced-round AES. The attacks demonstrate the capabilities of advanced models, but highlight the need for human verification and the limitations of current AI-powered cryptanalysis. AI summary
Firehose
Filtered to Hacker News, tagged “AI safety” · clear filters
Browse: People · Companies · Papers · Podcasts · Hacker News · Deep dives