OpenAI slowed development of its unreleased Astra model after it hit a critical cybersecurity threshold

OpenAI published preliminary cybersecurity evaluations for Astra, a model still in development, on August 7, 2026, along with the safeguards and security controls it is adding around it. TechCrunch reported the same day that the model reached what OpenAI calls its critical cybersecurity threshold — the point at which it could independently identify and carry out attacks against traditionally well-protected real-world systems — and that OpenAI slowed the model's development in response.

What it means A frontier lab publicly slowing a model on cyber-capability grounds is the clearest sign yet that capability thresholds now gate release schedules.