Encrypted AI reasoning vulnerability in Anthropic, OpenAI, and Google APIs let researchers decode 315,320 published session ...
Researchers found every major AI provider encrypts reasoning tokens with a single global key—and exploited it to decode a trove of data.
Replayable AI reasoning blocks let researchers recover API keys and passwords from public logs; they say the main extraction attack is now mitigated.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results