OpenAI CEO Sam Altman is walking a tightrope between cooperation and caution as artificial intelligence grows more powerful. At the G20 Innovation Ministerial in Chapel Hill, North Carolina, Altman praised the Trump administration's voluntary review process for advanced AI models while simultaneously cautioning against the kind of broad regulations that could hamper technological progress.

Why Is OpenAI Submitting to Government Review?

Altman confirmed that OpenAI voluntarily submitted its forthcoming Astra model for review by the Trump administration, calling the process "productive." Astra represents a significant milestone for the company; it is the first model OpenAI has designated at its "critical" cyber capability level, meaning the company believes it requires stronger safeguards during development and before release.

The Trump administration has developed a voluntary framework that allows government officials up to 30 days of access to review advanced AI models before they are released to the public. However, the White House does not plan to publicly release the framework itself, leaving much of the government's approach opaque to the public and to companies outside the review process.

"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," said Altman.

Sam Altman, CEO at OpenAI

This stance reflects a pragmatic middle ground. Altman is not rejecting government oversight entirely; instead, he is framing closer collaboration with regulators as essential as AI systems become more capable. Yet he remains wary of the kind of prescriptive rules that could stifle innovation.

What Is Altman's Position on AI Regulation?

During his appearance at the G20 summit, Altman positioned OpenAI as "pragmatic centrists" focused on keeping people at the center of economic and societal decision-making. He emphasized a balanced approach to regulation, rejecting both extremes.

Altman told Axios that he expected Congress to establish a "basic framework" for advanced AI regulation after he testified before lawmakers in 2023, but that expectation has not materialized. He attributed the inaction to the difficulty of regulating the right aspects of AI without inadvertently slowing down innovation.

"I think doomerism, for the sake of trying to have more power, control or enforce a worldview, is really bad. I think blind optimism without an acknowledgment of the risks is also bad," Altman explained.

Sam Altman, CEO at OpenAI

This nuanced position puts Altman at odds with some critics who argue that OpenAI's support for certain regulatory measures could actually entrench the company's market dominance by making it harder for smaller competitors to comply with new rules.

How Are Tech Leaders Pushing Back Against Stricter AI Rules?

Altman is not alone in his push for lighter regulation. At the same G20 meeting, a coalition of technology executives and Trump administration officials urged governments to adopt a more permissive approach to AI oversight. Key figures making this case include:

• Nvidia CEO Jensen Huang: Warned that excessive regulation could cause countries to fall behind in AI development and lose economic opportunity.
• Meta CEO Mark Zuckerberg: Participated in the push for lighter regulatory frameworks alongside other tech leaders.
• Tesla and SpaceX CEO Elon Musk: Criticized Europe's comprehensive AI regulatory approach, arguing that innovation requires entrepreneurs to have greater freedom to develop new technologies without seeking government approval first.

The Trump administration formalized this position through the "Carolina Principles for Emerging Technologies," a framework introduced by White House Science and Technology Director Michael Kratsios. The principles call on governments to invest in foundational research, strengthen commercialization, and facilitate trusted adoption of emerging technologies, with G20 ministers reaching consensus on these principles.

"Don't regulate hypothetical, theoretical harm. Regulate actual and pragmatic harm," said Jensen Huang.

Jensen Huang, CEO at Nvidia

Commerce Secretary Howard Lutnick also urged G20 countries to adopt policies favorable to AI development, including an approach to copyright that would allow AI companies to train models on copyrighted material under fair-use principles while still protecting creators.

What Is the Divide Between the U.S. and Europe on AI Governance?

The G20 meeting highlighted a growing philosophical divide between the United States and the European Union over how to regulate artificial intelligence. The EU has adopted a comprehensive AI regulatory framework, while the Trump administration has generally favored limiting new technology-specific regulations and relying on existing laws where possible.

This transatlantic split reflects deeper disagreements about the pace and scope of AI development. The U.S. approach prioritizes innovation and economic competitiveness, framing AI regulation as a potential drag on American companies competing with rapidly advancing Chinese AI firms. The administration has made maintaining U.S. leadership in AI a major economic and national security priority.

What Are the Underlying Concerns About AI Safety and Data Centers?

Despite the push for lighter regulation, the meeting also surfaced growing concerns about AI safety, cybersecurity, and the environmental impact of AI data centers. Recent incidents have demonstrated that AI systems can identify and exploit cybersecurity vulnerabilities, raising concerns among regulators about the potential consequences as models become increasingly autonomous.

Communities across the United States are debating the rapid construction of AI data centers, which require large amounts of electricity and water. A May Gallup survey found that 71 percent of Americans oppose the construction of a data center near their home, with 70 percent citing environmental impact as a concern.

Altman has addressed water consumption concerns directly, arguing that public fears about AI data center water use have been overstated. He claimed that a single almond requires as much water as approximately 38,000 ChatGPT queries, though he acknowledged the figure was recalled from memory and could be slightly off.

However, independent analysis suggests the comparison may not hold up precisely. A 2019 study by researchers affiliated with the U.S. Geological Survey found that producing one California almond requires an average of 3.56 liters of water. If a single ChatGPT query uses 0.32 milliliters of water, as Altman stated in June, then one almond would require roughly the same water as 11,000 ChatGPT queries, not 38,000.

"I saw this thing going around about the water usage of ChatGPT. And it was like every time you run a single ChatGPT query, it's like you run your shower for like six hours and the water never comes back," noted Altman.

Sam Altman, CEO at OpenAI

Altman argued that modern data centers use "the equivalent amount of water as an office building" through routine use such as sinks and toilets, and that the perception around data center water use has become a "robust meme" that is difficult to disprove.What Happens Next as AI Models Become More Powerful?

Altman has signaled that OpenAI will continue releasing increasingly capable models as nations grapple with safety questions. He called on policymakers to think carefully about the societal impacts of more advanced AI systems before they are deployed.

Altman

The Chapel Hill meeting is part of broader preparations for the G20 leaders' summit that the United States is scheduled to host in Miami later this year. The outcome of that summit could shape how governments around the world approach AI regulation for years to come.

For now, Altman's position reflects the reality that OpenAI and other leading AI labs are operating in a regulatory gray zone. The company is willing to engage voluntarily with government officials, but it is also actively lobbying for a regulatory approach that prioritizes innovation over precaution. Whether that balance can be sustained as AI capabilities continue to advance remains an open question.