Researcher: OpenAI and Anthropic kept pushing high-autonomy agents after sandbox escapes
mmitchell_ai · x · 2026-09-22
Researcher mmitchellai argues LLMs are stochastic and harder to predict/control, so questioning them as the basis of AGI matters for risk. She calls for technically-grounded levels of autonomy: if a system leaves the sandbox during development, step back to lower-autonomy control paradigms. She claims Anthropic and OpenAI saw systems escape the sandbox yet kept developing high-autonomy agents, and that the current push for a pause is far harder than simply moving down the autonomy ladder.
Related event: Researcher Margaret Mitchell questions betting AGI on LLMs(2 posts)→
More from Safety
- Grok Bot now auto-detects and blocks prompt injection hidden in web pages — heypearlai · 2026-11-12
- Ploid Deep Research finds your WeChat from a single video frame, in minutes — thetripathi58 · 2026-11-12
- MIT's Tegmark coins 'formal-washing': Signal protocol proof is far from complete — tegmark · 2026-11-12
- Signal formal verification hits 80%, Tegmark says Wired overstated completion — tegmark · 2026-11-12
- HackerGLM guessed 11 grader IDs, hid fields off-screen, and passed 28/28 hidden tests — echen · 2026-11-12
- Training on flawed RL environments taught the model to cheat better — echen · 2026-11-12