OpenAI: Astra Reaches 'Critical' Threshold in Cybersecurity Capabilities
nickbaumann_ · x · 2026-09-02
- OpenAI employee Romain Huet stated that Astra is significantly more token-efficient and capable at finding vulnerabilities and developing exploits.
- Officially, Astra has reached the "Critical" threshold under OpenAI's Preparedness Framework for cybersecurity.
- Access to its most advanced cyber capabilities will be limited at launch and expanded over time.
Related event: OpenAI Previews Astra as gpt-6-astra Traces Multiply Ahead of Launch(85 posts)→
More from Models
- Models can detect read-only probes with just last-two-layer SFT — Sauers_ · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Polymarket opens betting market on who releases the next Gemini Argon model — Polymarket · 2026-10-18
- Grok matter-of-factly lists which groups user was racist against, exempting Africans and Latin Americans — teortaxesTex · 2026-10-18