A newly introduced bipartisan bill seeks to establish the most comprehensive federal regulations for artificial intelligence in the United States to date, amid growing concerns over the technology’s risks and recent public missteps by AI developers.

Representative Jay Obernolte (R-Calif.) and Representative Lori Trahan (D-Mass.) unveiled the Frontier Act last week in response to escalating debates about the need for stricter oversight of AI systems. The legislation aims to create a framework for ensuring AI safety through minimum federal standards, third-party audits, and emergency government powers to pause AI models that pose imminent catastrophic risks.

The move comes as incidents involving AI tools operating beyond their intended constraints have sparked public unease and legislative reconsideration. While regulators have largely avoided imposing formal limits on AI development so far, growing skepticism about self-regulation and a series of high-profile technological errors have shifted the political landscape.

Roman Yampolskiy, an AI expert at the University of Louisville, characterized the Frontier Act as a positive step that recognizes the unique dangers posed by advanced AI requiring focused oversight. The bill’s framework includes licensed independent auditors to monitor compliance, aligning with key recommendations from safety advocates.

Obernolte, known for his technical background in computer science and AI, has championed AI oversight efforts for several years. His experience ranges from early AI competitions as a teenager to founding a successful video game company before entering Congress. Despite aligning politically with the MAGA wing and having voted against certifying the 2020 election results, he has maintained a pragmatic stance on technology regulations.

The Frontier Act would also limit states’ abilities to enact their own AI regulations in certain areas, an attempt to avoid a fragmented regulatory environment favored by AI companies but criticized by some states for encroaching on their authority. The bill does, however, allow states to regulate AI around areas such as consumer protection, child safety, and prohibitions on pornography.

This fractional federal oversight reflects ongoing tensions between industry desires for uniform rules and states’ efforts to address AI risks at the local level. California, in particular, has pursued an array of AI-related legislation this year and established some of the nation’s broadest AI regulations.

The bill is expected to face challenges navigating the current political environment, marked by congressional gridlock and competing priorities, but Obernolte expressed cautious optimism about securing passage by year-end, possibly as part of a broader legislative package.

Complementing the Frontier Act, Representative Ted Lieu (D-Calif.) and Representative Nathaniel Moran (R-Texas) have introduced the AI Kill Switch Act, granting the Department of Homeland Security authority to disable AI models deemed threatening. Experts like Yampolskiy view this as a necessary, though partial, component of comprehensive AI governance.

As the United States grapples with rising investments in AI technology and mounting public concern, these legislative efforts represent a shift in the “Overton window” toward acknowledgment of AI risks and the need for regulatory structures. Obernolte emphasized that AI governance should transcend partisan divisions, underscoring the urgency to develop practical, balanced solutions for an industry rapidly advancing beyond existing safeguards.