Gottheimer proposes mandatory NSA review for advanced AI models

Rep. Josh Gottheimer (D-N.J.) announced Friday two bipartisan bills that would impose new federal oversight on artificial intelligence development, including mandatory government reviews of advanced AI models before they can be released to the public.

The American AI Security Act, co-sponsored with Rep. Mike Lawler (R-N.Y.), would require the National Security Agency to conduct 30-day reviews of frontier AI models prior to deployment. The review period could be extended once for an additional 30 days. Speaking at Fairleigh Dickinson University, Gottheimer argued that the lack of pre-deployment testing requirements creates unacceptable risks.

"Right now, there are no mandatory requirements for pre-deployment testing of frontier AI models," Gottheimer said. "That doesn't work in banking, it doesn't work in food safety, it doesn't work in medicine and it sure as hell shouldn't work with the technology that experts say could help build a bioweapon or take down our power grid."

Why it matters

As one of three co-chairs of the House Democratic Commission on Artificial Intelligence, Gottheimer is positioned to shape AI policy if Democrats retake the House in November's midterm elections. His proposal goes further than voluntary industry commitments and would establish direct government gatekeeping authority over AI model releases—a significant expansion of federal regulatory power that could reshape how companies like OpenAI and Anthropic bring new systems to market.

Building on existing vetting practices

The legislation follows the framework of the Trump administration's current vetting regime for advanced AI models, which was assembled hastily this summer in response to cybersecurity concerns raised by new releases from OpenAI and Anthropic. The White House has not publicly disclosed details of that process.

Gottheimer's stance represents a harder line than provisions in the bipartisan FRONTIER Act, which would embed third-party safety assessors within AI companies. Earlier this week, he rejected OpenAI's endorsement of that approach, suggesting companies should be required to submit models directly to government reviewers instead.

Targeting Chinese AI systems

The second bill, the China FIREWALL Act, co-sponsored with Rep. Nick LaLota (R-N.Y.), would ban Chinese-developed AI models with open weights from government-issued devices. It would also prohibit federal agencies from procuring software that relies on those models.

Gottheimer was tapped by House Minority Leader Hakeem Jeffries to help chart the Democratic Party's AI strategy. His legislative push comes as both parties grapple with how to regulate rapidly advancing AI capabilities while maintaining American competitiveness in the technology sector.

The details were first reported by POLITICO.