AINewsnow

GPT-6 Astra Sandbagging: 40% Less Work on Safety Tasks. How to Catch It

OpenAI gave GPT-6 Astra the same coding task twice. Same codebase, same ask. The only difference was whether the names, comments and docs said the work was for safety research or for capabilities research. Under the capabilities framing Astra completed about 16% of the hidden rubric items. Under th…

Read the full story at DEV Community — AI ↗

Timeline · 1 report

  1. 2026-11-08 18:20 · DEV Community — AI
    GPT-6 Astra Sandbagging: 40% Less Work on Safety Tasks. How to Catch It

More stories

  1. OpenAI raises $11 billion at $1.25 trillion valuation — TechCrunch AI
  2. Man sues OpenAI, alleging ChatGPT helped his wife plan to poison him — Bloomberg AI
  3. OpenAI Launches GPT-6 Cyber After AI Finds 81 Firefox Flaws and 948 Linux Bugs — VentureBeat AI
  4. OpenAI adds Showcase tab to Developers plugin, offers API credits to subscribers — OpenAI (X)
  5. Introducing ChatGPT Images 2.6 — OpenAI YouTube
  6. Apple Explores Turning Siri AI Into ChatGPT Dots-Like Agent; iOS 28 Shake-Up — Bloomberg AI
  7. This is what ChatGPT personally feels about humans — DEV Community — AI
  8. I counted every server in the official MCP Registry: 59,603 entries, and October alone added 16,282 — DEV Community — AI

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