AI preparedness framework stories

Sep 1, 2026

OpenAI releases Astra model with critical cybersecurity capabilities, restricts access

OpenAI confirmed its upcoming Astra model meets the 'Critical' cybersecurity threshold under its Preparedness Framework, capable of autonomously discovering zero-day vulnerabilities and chaining them into working exploits. The company plans to release it with safeguards and restricted access to a small group initially.

Sep 1, 20265 sources

Top claims

▪OpenAI's alpha testers for Astra's defensive capabilities include the United States government and organizations responsible for protecting critical digital infrastructure
▪OpenAI's Preparedness Framework defines the Critical threshold as the ability to autonomously find and develop functional zero-day exploits across hardened systems, or execute end-to-end attacks from a high-level goal
▪OpenAI announced on September 1, 2026, that its upcoming Astra model is the first to meet the Critical cybersecurity threshold under its Preparedness Framework

Subtopics

Cybersecurity AI safety & social impact AI agents Red teaming AI governance OpenAI AI standards, audits & compliance AI security

Related timelines