OpenAI Pauses High-Risk Astra but Ships GPT-5.6-Cyber
eyishazyer · x · 2026-08-11
A user highlights that OpenAI made opposite calls on two highly capable cybersecurity models within the same week.
Key Events:
- Pausing Astra: On August 7, OpenAI paused the Astra model internally because it might hit their top cyber-risk tier. This model could independently find and chain zero-days against hardened targets.
- Shipping GPT-5.6-Cyber: On August 10, they shipped GPT-5.6-Cyber, which sits one risk rung lower. It answers 95% of exploit-development queries and has even found a real zero-day in Chrome's V8 engine (patched under a live CVE).
The author notes this shows OpenAI's safety framework working as intended, drawing a hard line at Critical while treating lower tiers as usable with access controls. However, it also implies that the capability "ceiling" companies are comfortable shipping is steadily creeping upward.
Related event: OpenAI Faces Internal Dispute Over AI Model Safety Ratings(2 posts)→
More from Models
- GLM 5.5 rumored to drop soon, coding ability expected to match frontier models — bindureddy · 2026-10-16
- Five rounds of adversarial prompting make Opus 5.5 and Argon converge on identical output — LanceCharger · 2026-10-16
- DIY benchmark ranks 18 models on subjective judgments, GLM-5.3 tops consensus match — arkuto · 2026-10-16
- Bindu Reddy: Google's Argon excels at complex doc writing at half the cost of Opus — bindureddy · 2026-10-16
- MAI-2 is 2x bigger than V4-Pro while V4.1F is 3x smaller: Microsoft's scaling ladder questioned — teortaxesTex · 2026-10-16
- Step 5 Preview hits DeepInfra day-zero: 600B MoE, 1M context, $1/M input tokens — StepFun_ai · 2026-10-16