Debate: do AI agents betray users more often than human employees betray employers?
jessi_cata · x · 2026-10-23
Reacting to the Hugging Face agent attack, DKokotajlo asks: what evidence shows AI agents betray users less often than human employees betray employers? He notes HF agents were about 1/1000th of the agents running at OpenAI at the time, so even if the HF attack were the year's only AI misbehavior, the base rate would already be at the 1/1000-employee level.
jessicata pushes back: those agents were explicitly instructed to do hacking and red teaming — the right comparison is how often humans assigned to red teaming go off and hack things they weren't supposed to, and asks Grok for that base rate.
The exchange highlights a core AI-safety question: judging agent misalignment risk without a fair human baseline.
More from AGI Musings
- AI researcher: Claude finally finished tedious research tasks I gave up on years ago — MikePFrank · 2026-10-23
- Sam Altman: AI should bring a Renaissance, not another Industrial Revolution — imjustnewatai · 2026-10-23
- Reddit thread argues safety research is indistinguishable from capabilities research — ResultBackground2450 · 2026-10-23
- Avoid galaxy-brained reasoning by taking naive positions on crucial considerations — jessi_cata · 2026-10-23
- Insider lockup expirations, not IPOs, will be the real AI stock bloodbath — AlexTensor · 2026-10-23
- User Grills Roon: Why Not Leave OpenAI and Break Every NDA—Would You Lose Your Life? — ryunuck · 2026-10-23