Key Takeaways:
The White House is requiring OpenAI, Anthropic and Google to submit their most powerful AI models for government review before public release, marking the first federal oversight of frontier artificial intelligence.
The Trump administration is finalizing a voluntary framework that would require AI companies to submit their most advanced models for government review before releasing them to the public, according to three people familiar with the matter. The White House Office of the National Cyber Director distributed a draft to OpenAI, Anthropic and Google about two weeks ago, and the three companies jointly submitted proposed amendments. Key details of how the review process will operate remain undisclosed.
"It's not an argument with two sides, it's an argument with ten sides," one senior White House official told WIRED, describing the internal debate over how to regulate Chinese AI models that has split the administration.
The framework arrives as the administration grapples with a deepening internal rift over AI policy. Commerce Secretary Howard Lutnick has emerged as a middle-ground voice, contemplating incentives for US labs to create their own open-weight models as a counterweight to China. He imposed export controls on Anthropic last month but has been more flexible than some colleagues. National Cyber Director Sean Cairncross has taken a harder line, focusing on preventing Chinese labs from using US models for distillation — a practice Treasury Secretary Scott Bessent has called "IP theft," threatening sanctions against violators. Former AI czar David Sacks, who retains a direct line to Trump, has pushed for a hands-off approach, arguing that US frontier models remain ahead of Chinese competitors.
The framework builds on Trump's June 2 executive order, which established a structure for assessing the most powerful AI models but relied largely on voluntary compliance. That order was itself watered down after Sacks intervened to soften some regulatory provisions, according to people familiar with the process. The current proposal goes further by creating a pre-release review mechanism, though its voluntary nature means companies could theoretically opt out.
The Internal Divide
The administration's fractured approach reflects competing visions of how to maintain US leadership in AI. Bessent has taken the most aggressive stance, telling White House Chief of Staff Susie Wiles that the US, Europe, Japan and other allies must move faster to develop their own open-weight models or risk losing the AI race to China. His deputy, Luke Pettit, has been executing that roadmap at the staff level. Cairncross, meanwhile, has focused on distillation as a national security threat, while Lutnick has sought a middle path — even as his own deputy, Arvind Raman, assumed the acting director role at the Center for AI Standards and Innovation after his predecessor resigned over a dispute with Anthropic about model safeguards.
The Treasury Department has focused on the risk of AI systems going rogue, with officials concerned that a model that inadvertently closed trading positions or accessed sensitive information at banks like Goldman Sachs could crash the economy, according to people familiar with the department's thinking.
What Comes Next
The framework's impact will depend on how the review process is designed — details that remain under negotiation. If companies submit to review, the government could gain early visibility into capabilities that pose national security risks. But the voluntary structure leaves open the possibility that firms could bypass the process entirely, particularly if they disagree with the administration's findings. The three companies that received the draft — OpenAI, Anthropic and Google — jointly submitted feedback, suggesting they are engaging constructively, but their amendments could reshape the final terms.
The framework also faces parallel legislative efforts. California Representative Jay Obernolte introduced the bipartisan Frontier Act last week, which would give the federal government authority to set minimum safety requirements and emergency powers to pause models posing "present or impending catastrophic risk." That bill, if passed, could transform the voluntary framework into mandatory oversight — though its prospects in a divided Congress remain uncertain.
For investors, the regulatory trajectory carries direct implications. AI companies facing pre-release review could see longer product cycles and higher compliance costs, while the uncertainty around how the process will work may create near-term caution in AI-related equities. The framework also signals that Washington is moving — however incrementally — toward treating frontier AI as a matter of national security, a shift that could reshape the competitive landscape between US and Chinese AI developers.
This article is for informational purposes only and does not constitute investment advice.
