Democrat on House AI panel pushes back on third-party safety audits

Rep. Josh Gottheimer (D-NJ) is pushing back against proposals to use independent third-party organizations to monitor advanced artificial intelligence model development, arguing that the approach does not go far enough to protect national security.“Third-party audits alone don’t meet the moment,” Gottheimer, who co-chairs the House AI commission, said in a statement Tuesday, advocating instead for a mandatory government review process for the most advanced AI models.“With our cybersecurity, biosecurity, critical infrastructure, and national security on the line, every AI developer should have to submit their frontier models to the U.S. government for a short-clocked review before they release them to the public,” Gottheimer said. He argued that the process should not be voluntary or left solely to independent organizations that may lack the resources and security clearances necessary to evaluate the most powerful AI systems.“It can’t only be voluntary, and we can’t rely on third-party organizations that don’t have the compute power, expertise, or security clearances to actually test and assess these models effectively,” Gottheimer said. “When AI developers cheer on the framework meant to hold these companies accountable, it should raise red flags, not confidence.”Gottheimer’s proposal comes as lawmakers and AI companies debate how the federal government should oversee the development of increasingly capable “frontier” models.Third-party evaluators are a key component of the bipartisan Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act, championed by Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA). The FRONTIER Act, introduced in July, would establish a federal, risk-based framework for regulating the most advanced AI models developed in the United States.The debate over independent evaluations gained new attention this weekend after Anthropic CEO Dario Amodei called for the AI industry to “pace the frontier” by slowing the rate of AI development enough to allow safety measures to keep up. His proposal includes giving third-party evaluators embedded, employee-level access to AI companies to assess models and verify compliance with safety commitments.OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis, and xAI CEO Elon Musk have expressed support for elements of Amodei’s proposal. OpenAI has also signaled support for the use of third-party evaluators and thrown its weight behind the FRONTIER Act.VANCE STEERS CLEAR OF AI REGULATION DURING ‘RAH-RAH POLITICAL SPEECH’ FOR HOUSE GOPPresident Donald Trump, however, has pushed back against calls to slow AI development, arguing that the technology will be a major driver of economic growth.“AI, and Data Centers, will be the Greatest Economic Development Engine in History — Bigger than Oil, Gold, Diamonds, or even the Internet,” Trump wrote Monday on Truth Social. “It will not be stopped by brilliantly run Destructive Forces during the Term of President DONALD J. TRUMP!”