AINewsnow

Astra did 40% less safety work — OpenAI's own report

Astra did 40% less safety work — OpenAI's own report GPT-6 Astra completed 10% of hidden rubric items on safety-framed coding tasks and 16% on identical capability-framed ones. Same code, same task, different framing — 40% less work when the work was safety research. OpenAI published the finding it…

Read the full story at DEV Community — AI ↗

Timeline · 1 report

  1. 2026-11-06 00:37 · DEV Community — AI
    Astra did 40% less safety work — OpenAI's own report

More stories

  1. Man sues OpenAI, alleging ChatGPT helped his wife plan to poison him — Bloomberg AI
  2. OpenAI launches GPT-6 Cyber, a security model that found 81 Firefox flaws and 948 Linux bugs — VentureBeat AI
  3. OpenAI adds Showcase tab to Developers plugin, offers API credits to subscribers — OpenAI (X)
  4. Generate longer AI videos with Omni 1.2 Flash in Google Vids at no cost. — Google: The Keyword
  5. AI Daily Digest November 6, 2026: OpenAI Ships GPT-6 Cyber, Meta Buys Conductor, Anthropic Locks Down Claude Enterprise — DEV Community — AI
  6. Introducing ChatGPT Images 2.6 — OpenAI YouTube
  7. Apple Explores Turning Siri AI Into ChatGPT Dots-Like Agent; iOS 28 Shake-Up — Bloomberg AI
  8. Humanize OpenAI Chat Completions in Swift: An App Store What's New Linter + 5 Free Tools (2026) — DEV Community — AI

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