The proposed legislation, still in its formative stages, would specifically preempt state requirements that compel top AI developers to disclose information about new models to identify critical safety or security risks.
The talks, led by Rep. Jay Obernolte (R-Calif.) and Rep. Lori Trahan (D-Mass.), represent the latest attempt to craft federal rules governing the rapidly evolving technology. According to two tech lobbyists and three AI policy advocates who spoke on condition of anonymity due to the sensitivity of the negotiations, the bill would not preempt state laws that do not directly regulate model developers. This narrow focus suggests an effort to carve out a federal standard for the most advanced systems while leaving other state-level AI regulations intact.
A central point of contention has emerged between the two lawmakers. Three of the people familiar said Obernolte favors a light-touch or even voluntary approach, allowing AI companies to decide whether to disclose certain information to the government. Trahan, by contrast, is pushing for greater accountability, including mandatory data-sharing requirements for developers of cutting-edge models. The disagreement underscores a broader ideological divide over how aggressively the federal government should police the industry.
Sunset Provision and the Mythos Challenge
The negotiators are also discussing a sunset provision that would allow states to reclaim authority over frontier AI development after two years, according to four of the people familiar. This temporary preemption would give Congress a defined window to establish a permanent federal framework before states can reassert their own regulations. The talks follow several failed bids to reach consensus in Congress over AI guardrails, highlighting the difficulty of legislating a technology that evolves faster than the legislative process.
The new details of the House negotiations come as the White House grapples with similar questions posed by the emergence of Mythos, a powerful new model developed by top AI firm Anthropic. The administration is facing pressure to define its own approach to federal oversight, even as lawmakers on Capitol Hill attempt to set a national standard. The parallel efforts between the executive and legislative branches could create friction or, potentially, a unified strategy for governing the most advanced AI systems.
Proponents of the federal preemption argue that a patchwork of state laws would stifle innovation and create compliance chaos for developers operating nationwide. Critics, however, warn that a weak federal standard could gut the most stringent state protections, leaving consumers and national security interests exposed. The outcome of the Obernolte-Trahan talks is likely to shape the broader debate over who holds the reins on one of the most consequential technologies of the century.
