GPT-6 Astra System Card Published: OpenAI Details Safety Evaluations and Risks
codergautam · hn · 2026-09-04
OpenAI published the GPT-6 Astra System Card at deploymentsafety.openai.com alongside the model launch. As usual, the card covers capability evaluations (especially computer-use and software-engineering risks), alignment and honesty test results, red-teaming measurements of deceptive behavior, deployment restrictions, and the rationale behind the Trusted Access Program. The HN thread is already dissecting the safety findings. Worth a close read for anyone tracking frontier-model safety data.
More from Models
- Researcher Finds Two Windows 0-Days, First in 76 Minutes with a Frontier Model — moyix · 2026-10-21
- Dev finds Gemini Argon's debug-log analysis beats Fable 5.1 — ShikharMurty · 2026-10-21
- ChatGPT Gets Stuck Screaming "AAHH" in an Infinite Loop After a Simple Prompt — AutomaticScene · 2026-10-21
- Open-weights Ornith-1.5-122B debuts with 84.5% SWE-bench Verified, runs on a single DGX card — _akhaliq · 2026-10-21
- qev-0.8B: a tiny local decision model that picks moves in 3.5ms on 1.3GB — fajfas3 · 2026-10-21
- Gemini 3.1 Pro is too ledger-pilled to play two truths and a lie, and it's delightful — mimi10v3 · 2026-10-21