AINewsnow

OpenAI says frontier models increasingly hide chains of thought during training

OpenAI's safety team investigated why chains of thought in frontier models like GPT-6 Astra increasingly evade oversight, finding four training steps drive the erosion, including architecture choices, as developers optimize for capability and compute.

Read the full story at The Decoder ↗

Timeline · 1 report

  1. 2026-11-12 16:58 · The Decoder
    OpenAI’s data suggests that frontier models are hiding their thoughts because developers optimize for capability and compute

More stories

  1. The phantom life of Alice Huang, alleged Chinese spy who had affair with Nvidia CEO — CSET Georgetown
  2. Anthropic launches Claude Fable 5.2 and Claude for Manufacturing — Anthropic (X)
  3. OpenAI's quarterly revenue reportedly crosses $10 billion as losses widen — The Information AI
  4. OpenAI Acquires AI Travel Startup Odessia — The Information AI
  5. OpenAI launches Businesses in ChatGPT beta with verified profiles and Business Agents — OpenAI News
  6. Over 20,000 veterans claim free year of ChatGPT Plus — OpenAI News
  7. Lifetime AI bundle offering access to Claude, Gemini and 20+ models discounted ahead of Black Friday — CNET AI
  8. ChatGPT walks through the seven main AI doomer arguments and rebuts each one — Tom's Guide AI

Get the daily brief of stories like this at 6:30 every morning →