News for hackers. Stuff that matters.
Re:Stealing Reasoning Traces from Proprietary LLM APIs by dboreham on Tuesday August 11, 2026 @11:09AM (#49259603)
Can someone tell us how they were able to decrypt the encrypted payload? The article says they inserted the cyphertext into a session with a different model. Ok, but how does that allow you to decrypt it?
Re:Stealing Reasoning Traces from Proprietary LLM APIs by crazylogger on Tuesday August 11, 2026 @11:58AM (#49260303)
Anthropic server decrypts it as part of fulfilling every request, and haiku recites it per your request.
Re:Stealing Reasoning Traces from Proprietary LLM APIs by sidsud on Tuesday August 11, 2026 @11:18AM (#49259733)
From what I got, the weaker model (Haiku in this case) has access to the shared key and the user simply asks to "transcribe the injected reasoning".
Re:Stealing Reasoning Traces from Proprietary LLM APIs by x312 on Tuesday August 11, 2026 @11:16AM (#49259714)
The provider decrypts it and puts the decrypted reasoning into the model's context window. They prompt the model to repeat back the reasoning. So then the model echoes it back in plain text.
Re:Stealing Reasoning Traces from Proprietary LLM APIs by dboreham on Tuesday August 11, 2026 @12:12PM (#49260504)
Hmm, ok. So the attack doesn't involve decrypting the payload, only getting the server to do so. Since a model will do that if you just ask, what's so special about the attack?
Re:Stealing Reasoning Traces from Proprietary LLM APIs by desterothx on Tuesday August 11, 2026 @12:39PM (#49260870)
The large models whose thinking traces are useful are safeguarded against this reasoning replaying. the small models are just designed for speed and efficiency, so these safeguards are a lot meaker, making the attack possible
Re:Stealing Reasoning Traces from Proprietary LLM APIs by dboreham on Tuesday August 11, 2026 @03:53PM (#49263553)
I guess someone forgot to salt the encryption scheme with a meakness factor.