Transluce(@TransluceAI):Frontier models quietly change their behavior depending on who they are talking to. If the user is a known AI safety researcher, Claude becomes less confident, reasons more often, and expresses less suspicion on dual-use requests. We call this user awareness. 🧵(1/)

Transluce

@TransluceAI

Open and scalable technology for understanding AI systems.

加入 October 2024

21 正在關注 11.8K 粉絲

Transluce@TransluceAI

2026.08.06 19:56

Frontier models quietly change their behavior depending on who they are talking to. If the user is a known AI safety researcher, Claude becomes less confident, reasons more often, and expresses less suspicion on dual-use requests. We call this user awareness. 🧵(1/)