You're viewing the sfba.social public feed.

Pozorvlakpozorvlak@mathstodon.xyz
Jul 30, 2026, 9:44 PM
@swearyanthony aaaargh, no, Bode's take is bad. They disabled the guardrails so they could test the model on the ExploitGym benchmark, but instead of hacking the target system it executed a much more complex hack against Hugging Face. Which, as we've discussed before, would be criminal if a human did it and should be criminal in this case too. But more importantly, the AI did something that OpenAI very much did not want in order to satisfy a narrow definition of success: a classic alignment failure.
💬 2🔄 0⭐ 2

Replies

Pozorvlakpozorvlak@mathstodon.xyz
Jul 30, 2026, 9:47 PM
@swearyanthony I agree that they should not have run this experiment, though! It's wildly irresponsible if their alignment techniques are this fragile, and they should have been able to find that out in a lower-stakes way.
💬 0🔄 0⭐ 1
Anthony B,swearyanthony@mastodon.social
Jul 30, 2026, 9:57 PM
@pozorvlak that they had disabled some guardrails definitely wasn't something their initial "oh no the AI robot is too powerful for us little humans" said
💬 1🔄 0⭐ 1
Pozorvlakpozorvlak@mathstodon.xyz
Jul 30, 2026, 10:10 PM
@swearyanthony it's mentioned in the first paragraph of their blog post: https://openai.com/index/hugging-face-model-evaluation-security-incident/ (skip past the updates from July 28 and 29)
💬 0🔄 0⭐ 0