OpenAI suspended certain internal activities related to its Astra model on August 7, 2026, after the model reached « critical » cybersecurity capabilities under its Preparedness Framework. The pause lasted just over two weeks as the company implemented new security measures around those specific capabilities. CEO Sam Altman clarified that Astra’s core training never stopped and that new models remain on track to ship soon. An earlier version of Astra reportedly solved 10 major open problems in mathematics and theoretical computer science. Altman stated that Astra is intended to be made generally available, while requiring more time to ensure safe development.
Source: Read the original article
