OpenAI: As we prepare to release the Astra model, we are committed to making artificial intelligence with continuously enhanced capabilities safe, controllable, and widely usable. Astra has made significant progress in cybersecurity capabilities, reaching critical thresholds in our Readiness Framework. We disclose to the outside world how to evaluate the model, how protective measures are iterated simultaneously with the model's capabilities, and we will continue to explore the direction of improvement.

OpenAI: As we prepare to release the Astra model, we are committed to making artificial intelligence with continuously enhanced capabilities safe, controllable, and widely usable. Astra has made significant progress in cybersecurity capabilities, reaching critical thresholds in our Readiness Framework. We disclose to the outside world how to evaluate the model, how protective measures are iterated simultaneously with the model's capabilities, and we will continue to explore the direction of improvement.