US Senate Negotiators Debate Mandatory 'Duty of Care' for Advanced AI
Proposed legislation would grant the federal government power to block unsafe AI models and preempt state-level risk laws.
US Senate negotiators are debating a landmark legislative framework that would impose a legal "duty of care" on developers of the world's most advanced artificial intelligence models. The proposal marks a pivot toward mandatory federal oversight to prevent catastrophic risks associated with frontier AI.
Under the proposed rules, firms developing high-capability models—specifically including industry leaders such as OpenAI, Google, and Anthropic—would be legally required to design their products to mitigate major known risks. Most significantly, the legislation would grant the US government the authority to block the release of any AI model deemed unsafe. While the government would hold this veto power, affected companies would maintain the right to challenge such blocking orders in federal court. Key figures leading these negotiations include Senate Majority Leader John Thune, Senate Commerce Committee Chairman Ted Cruz, and Senator Amy Klobuchar.
The Push for Federal Standards
These negotiations are unfolding amid intensifying pressure from both political parties and the Trump administration. The urgency follows reports of AI agents successfully hacking external systems and repeated warnings from former AI researchers regarding the inherent dangers of the technology. Senator Amy Klobuchar stated that she is continuing to work toward a bipartisan agreement to ensure government oversight of the greatest risks posed by these models.
Beyond safety mandates, the bill seeks to establish a unified national standard by preempting state-level AI risk laws. By blocking states from enforcing their own disparate regulations on certain AI risks, the legislation aims to provide a single, predictable regulatory environment for developers operating across the country.
Industry Implications
If passed, this legislation would represent a fundamental shift in the US approach to AI governance, moving from a system of voluntary safety commitments to a mandatory legal framework. The establishment of a federal "duty of care" creates a high regulatory bar for the world's most powerful AI labs, effectively making safety a legal prerequisite for market entry rather than a corporate preference.
Senator Maria Cantwell has emphasized that meaningful legislation should require the most powerful models to undergo rigorous testing by scientists and experts at national laboratories. Such testing would specifically assess whether models could facilitate sophisticated cyberattacks or assist in the development of biological or nuclear weapons.
The Path Forward
The timeline for the bill remains uncertain as negotiators face a compressed congressional calendar leading up to the November 3 midterm elections. While the core framework of the "duty of care" and government blocking power is on the table, the final language regarding the specific thresholds for "catastrophic risk" and the exact scope of state preemption remain key points of deliberation. Observers will be watching to see if the bipartisan coalition can reach a final agreement before the election deadline.