OpenAI says Astra reaches its toughest safety threshold yet

OpenAI says its next model, Astra, is capable enough to trigger the company's highest safety threshold to date.
The classification means additional safeguards and deployment controls are required before release, reflecting increased concern about the potential capabilities and misuse risks of more advanced AI systems.
The development highlights how frontier-model progress is increasingly being constrained not only by compute and performance, but also by safety and governance requirements.