# @davidad — 2024-12-07

♥106 ↻7 · https://x.com/davidad/status/1865478052995620969

“The *LLM* isn’t situationally aware, deceptive, or sandbagging—that’s silly anthropomorphism. It’s just that when evals (or people) test it, there are contextual cues of testing that prompt it to *roleplay* as ‘an AI being safety-tested’—an archetype which is often deceptive,” https://t.co/ggVbVWRvxn

tags: author:davidad, kind:tweet, on:observations, year:2024
cited on: observations