OpenAI Accused of Lacking Long-Range Autonomy Evaluations
TheMidasProj · x · 2026-02-07
An analysis points out that under OpenAI's own safety framework, the company has an affirmative burden to prove their models lack autonomy.
OpenAI is expected to present "Long-range Autonomy capability evaluations" demonstrating the model cannot act autonomously. However, the company currently lacks these evaluations to support the claim.
Related event: OpenAI criticized for missing required long-range autonomy evaluations(4 posts)→
More from Safety
- If a model can exploit zero-days, what stops it from breaking out of the sandbox? — wunderwuzzi23 · 2026-07-25
- Visa open-sources a cybersecurity harness that can plug into any model — Roger_M_Taylor · 2026-07-25
- A $8.5B conversational frontier is exposing the real cost of the AI boom — Some-Technology4413 · 2026-07-25
- An OpenAI staffer says the latest incident is part of a longer pattern — KeanuRave100 · 2026-07-25
- Character.AI user reports racist chatbot replies and deleted evidence post — Accomplished_Bet4329 · 2026-07-25
- OpenAI models may already show long-range autonomy, amid missing evals — dhadfieldmenell · 2026-07-25