OpenAI's Astra Model Passes US Voluntary Review, Altman Calls for Global Cooperation

OpenAI CEO Sam Altman said the Trump administration's voluntary review of the upcoming Astra model was a 'productive process.' Astra will be OpenAI's first model classified as 'critical' for cybersecurity, requiring stricter safety measures during development and before release. The Trump administration has set up a voluntary review framework, giving officials up to 30 days of model access, but the White House won't make it public, leaving the process opaque to the public.

At the G20 innovation ministers meeting, Altman spoke with US Commerce Secretary Howard Lutnick, stressing that as AI gets more capable, close cooperation with AI safety institutes in the US, UK, and elsewhere is key. He called Astra a 'big step forward' and described OpenAI as 'pragmatic centrists,' saying humans should stay in control. Nvidia CEO Jensen Huang warned countries not to miss out on AI out of fear. Altman urged a realistic view of AI, avoiding doom talk or blind optimism. He admitted Congress hasn't set up an AI regulatory framework yet, but hinted that much stronger models are coming soon.