Sam Altman: OpenAI's Astra Too Powerful For Release Now

OpenAI CEO Sam Altman says their new Astra model is too powerful to release right now. OpenAI is focusing on making Astra safer before a public launch. What's interesting is that frontier AI development continues to advance quickly. Just weeks after GPT-5.6 Sol, OpenAI now has Astra, an even more powerful model. Altman stated that OpenAI has no plans to restrict Astra, aiming to make it generally available rather than keeping it to a select few. However, Altman also noted that given Astra's cyber capabilities, more time is needed to ensure its safety. OpenAI's internal evaluations show significant advances in agentic coding and cybersecurity with Astra. The company is using its Preparedness Framework to guide this process. This framework helps identify capabilities and decide on necessary steps as new levels of AI power emerge. The bottom line is that while AI models are becoming incredibly capable, safety remains a critical concern before they can be widely deployed.