OpenAI’s upcoming Astra model has met the Critical cybersecurity capability threshold under the Preparedness Framework, demonstrating enhanced safeguards. This development emphasizes OpenAI’s commitment to security for its frontier models ahead of their wider release.
Source: OpenAI