Gary Marcus warns OpenAI may cross AI safety redline
GaryMarcus · x · 2026-09-02
Gary Marcus issues a "Red Alert" warning that OpenAI might be crossing a critical AI safety redline. Citing reports that OpenAI is exploring techniques to conceal model "thinking," Marcus argues this could make Chain of Thought (CoT) monitoring difficult or impossible. He quotes former OpenAI safety staff Steven Adler, stating that limiting CoT visibility violates one of the few redlines in the industry.
Related event: OpenAI's Reported Move to Hide Chain-of-Thought Sparks AI Safety Firestorm(11 posts)→
More from Companies & People
- Developers tire of frequent model switches, prefer stable improvements like Claude Code — ivan_bezdomny · 2026-09-02
- Report: OpenAI Broke Safety Taboo with Astra Model, Escalating AI Race — GarrisonLovely · 2026-09-02
- SEO Case Study: Sudden Traffic Growth via Keyword Mining — mazzaTalk · 2026-09-02
- Geoffrey Irving sad about OpenAI restarting large training run — geoffreyirving · 2026-09-02
- VCs Should Value Authenticity Over AI-Generated Pitches — jefrankle · 2026-09-02
- Reports of looped transformer progress from unexpected Chinese players — teortaxesTex · 2026-09-02