GPT-6 Astra system card: first OpenAI model to hit Critical cybersecurity threshold
RyanGreenblatt · x · 2026-09-06
OpenAI's GPT-6 Astra system card reveals the model is the first to reach the Critical cybersecurity level under its Preparedness Framework—able to find unknown vulnerabilities and develop exploits unaided with proper tools. OpenAI strengthened safeguards: stricter isolation, checkpoint encryption, full-trajectory monitoring including CoT, and blocking alignment evals before internal use. New robustness training makes Astra far more jailbreak-resistant than GPT-5.6 Sol, with tightened refusal boundaries for high-risk users plus regression testing and automated red-teaming. Ryan Greenblatt praises the monitorability/alignment detail but flags eval limits: limited elicitation and meta-gaming uncertainty.
More from Models
- Meta researcher: horrifying infra, not talent, is why OSS can't catch closed frontier — A_K_Nain · 2026-10-23
- TypeSafe's Jev: a "Decision Model" LLM That Returns Floats at $0.042/M Input Tokens — ariG23498 · 2026-10-23
- Microsoft's MAI-Base-2 scaling ladder pegs the L42 run at ~1.8×10^23 FLOPs — teortaxesTex · 2026-10-23
- Local model agentic coding benchmark: 27B matches 125B, tool-calling is the lever — Joralk_ · 2026-10-23
- NVIDIA open-sources PersonaPlex-7B, a full-duplex voice AI model that listens and talks simultaneously — solyarisoftware · 2026-10-23
- NVIDIA's KumoRFM-2: a foundation model that predicts churn, fraud and demand straight from relational databases — solyarisoftware · 2026-10-23