Bipartisan Ai Bill Is Getting Preempted By Bipartisan Resistance

Efforts to enact comprehensive federal artificial intelligence (AI) regulations keep getting preempted. In the most recent example, a bipartisan discussion draft of the Great American Artificial Intelligence Act is encountering opposition from both Democrats and Republicans, underscoring how disagreement over federal preemption of state AI laws has emerged as one of the biggest obstacles to enacting a national framework.

According to a Monday (July 6) analysis by Covington & Burling, the draft legislation released by Reps. Jay Obernolte, R-Calif., and Lori Trahan, D-Mass., has drawn criticism from opposite ends of the political spectrum over its treatment of state AI regulation. Democrats argue the bill goes too far by restricting states’ authority to regulate advanced AI model development, while many Republicans and industry stakeholders contend it does not go far enough because it leaves intact most state regulation governing AI deployment and applications.

The dispute illustrates the increasingly difficult balancing act facing Congress as it attempts to establish a national AI framework while dozens of states continue to enact their own AI laws. The preemption debate has become the central fault line in federal AI policymaking, raising questions about whether Congress can reach consensus on a nationwide regulatory regime.

The Obernolte-Trahan discussion draft would establish federal disclosure, transparency and risk-mitigation requirements for developers of frontier AI models while assigning oversight responsibilities to the Center for Artificial Intelligence Standards and Innovation (CAISI) at the National Institute of Standards and Technology. Unlike broader preemption proposals advanced earlier this year, however, the measure adopts a narrower approach to state regulation.

Under the proposal, states would be barred for three years from enforcing laws that specifically regulate the development of AI models. At the same time, states would retain authority to regulate activities occurring after deployment, including AI applications such as companion chatbots, as well as laws of general applicability. As a result, the legislation would preempt regulation of model development while preserving state oversight of many downstream AI uses.

Last year’s failed reconciliation legislation would have imposed a 10-year moratorium on state and local AI regulation, while a subsequent proposal by Sen. Ted Cruz, R-Texas, would have conditioned federal broadband funding on states pausing AI regulation. The Senate overwhelmingly rejected that proposal by a 99-1 vote, leaving the final legislation without any AI preemption provision. The current draft also is less sweeping than President Donald Trump’s December executive order directing the Justice Department to challenge state AI laws deemed inconsistent with administration policy.

Even the more limited preemption language has proven politically contentious. Some House Democrats argue that preventing states from regulating frontier AI development would leave important gaps in oversight while Congress continues to struggle to enact comprehensive legislation. Rep. Ted Lieu, D-Calif., who has separately worked with Obernolte on AI legislation, criticized the new proposal as failing to address concerns raised by civil rights organizations, labor groups and watchdog organizations. The proposal has also generated opposition from state lawmakers concerned about preserving their regulatory authority.

At the same time, many industry groups have not embraced the proposal despite continued expansion of state AI laws. According to the analysis, some stakeholders favor broader federal preemption to replace what businesses increasingly describe as a fragmented patchwork of state requirements.

Beyond the preemption debate, the Obernolte-Trahan legislation would establish extensive compliance obligations for developers of the largest frontier AI models. The requirements largely mirror provisions already adopted or proposed in states including California, Connecticut, New York and Illinois. Developers meeting specified computing thresholds would be required to publish frontier AI safety frameworks, issue transparency reports describing model capabilities and intended uses, undergo semiannual third-party audits by CAISI-licensed organizations.

The Covington analysis concluded that the legislation is unlikely to advance in its current form during the remaining months of the 119th Congress. Nevertheless, the authors argued that continued advances in frontier AI models, combined with the steady expansion of state AI regulation despite the prospect of federal legal challenges, will likely increase pressure on lawmakers from both parties to eventually reach a compromise on preemption.

The post Bipartisan AI Bill Is Getting Preempted by Bipartisan Resistance appeared first on PYMNTS.com.