CHAPEL HILL, N.C. — OpenAI Chief Executive Officer Sam Altman described the Trump administration’s voluntary review of the company’s upcoming Astra model as a “productive process” and said that closer engagement with government safety institutions around the world will grow more important as AI systems become more capable.
Why this matters
OpenAI says Astra is its first model designated at a “critical” cyber capability level, prompting the company to adopt stronger safeguards during development and before public release. The Trump administration has created a voluntary framework for reviewing certain advanced AI models prior to release that could include giving government officials access to models for up to 30 days. The White House does not plan to make that framework public.
Because the process is voluntary and not fully public, much of the government’s approach remains opaque to outside companies and the public even as leading labs participate and more powerful models are deployed.
Altman’s remarks
In an interview with Axios, Altman confirmed that the administration reviewed Astra and emphasized the voluntary nature of the process: “But we of course did it.” He added that “as these models get to a quite significant level of capability, I think the importance of really doing this, closely engaging with the safety institutes in the U.S., the U.K. and elsewhere in the world, will become more important.”
Altman described Astra as “a significant step forward.”
Context at the G20 ministerial
Altman traveled to the G20 Innovation Ministerial on Wednesday to discuss AI policy and met with commerce and technology ministers as the United States pitched its case to lead in AI. During a fireside chat at the summit lunch, Altman framed OpenAI as “the pragmatic centrists,” saying people should remain at the center of economic and societal decision-making and that “people should run the future, not AI.”
At a summit breakfast, Jensen Huang, chief executive of Nvidia, warned against fear-driven messaging about AI, telling leaders that the worst outcome would be failing to take advantage of AI and being left behind. Huang’s comments drew agreement from many ministers in the room.
Altman called for a “realistic” and “accurate” approach, criticizing both doomerism used to gain power or control and blind optimism that ignores risks.
Criticisms and the regulatory picture
OpenAI and Anthropic have faced criticism that their support for some forms of AI regulation could advantage large labs and raise barriers for smaller competitors. Altman said he had expected Congress to establish a “basic framework” for advanced AI after he first testified before lawmakers in 2023, but that has not yet occurred.
He attributed the inaction in part to the rapid pace of technological change and the difficulty policymakers face in regulating the right aspects without hindering innovation, noting that it “appears to be very hard to get Congress to do big things.”
What’s next
Altman expects OpenAI to continue releasing models while nations wrestle with safety questions. “Take our word for it that we have much, much, much more capable models coming soon, and try to think about what all the impacts of that on society could be and then need to get that right,” he said.
The remarks underscore that ongoing dialogue between industry and governments, and international cooperation on safety institutions, will remain central as more powerful AI models are developed and released.
