OpenAI pauses Astra work after internal test flags critical cyber‑attack capabilities
OpenAI announced that its in‑development Astra model hit the “critical cybersecurity threshold,” meaning it could autonomously discover and exploit zero‑day vulnerabilities in hardened systems. Because of this risk, the company has halted non‑essential work on Astra and is tightening testing environments, monitoring, and encryption to prevent misuse.