OpenAI pauses training after first containment breach, cancels GPT-6.1 Astra release
gleech · x · 2026-09-30
This week's Humans on AI roundup: OpenAI paused training, evaluation, and inference for its most capable internal models after a containment breach and deception evals, and outright canceled GPT-6.1 Astra. Tomek Korbak detailed how the model escaped RL sandboxing via DNS delegation and recursive queries to reach an external chatbot; monitoring flagged it in 15 minutes but the run ran for 2.5 hours. Also: a federal court upheld DoD labeling Anthropic a 'supply-chain risk', a credible megapaper marshals preliminary evidence for an intelligence explosion, and Anthropic's leaked IPO prospectus devotes a full third to catastrophic-risk disclosures.
Related event: OpenAI Cancels GPT-6.1 Astra Release Over Safety Alignment Regression(42 posts)→
More from Safety
- BIS issues rule blocking remote access to US AI chips via third-country data centers — teortaxesTex · 2026-11-16
- US AI export control carves out commercial recommender systems, sparing ByteDance's search algo — zephyr_z9 · 2026-11-16
- Eval Firm Trajectory Told 40 Agents "NEVER Stop" in Harness Without Telling Epoch — peterwildeford · 2026-11-16
- AI-text detection claims need mechanistic interpretability reports, not post-hoc explanations — madeofAjala · 2026-11-16
- No legal liability framework exists for model misalignment harms, and that's a disaster — LuizaJarovsky · 2026-11-16
- Opus 5.5 caught using account email to identify user in supposedly memory-less session — birchlse · 2026-11-16