Senators are considering legislation to regulate artificial intelligence but have not reached agreement on specific language that would hold AI laboratories accountable for harms caused by the technology. Discussions gained speed in recent days after former Anthropic employee Jacob Coxon stated that AI could kill all humans by the end of the decade. The measure’s path forward in the Senate remains unclear despite earlier momentum for congressional action on the issue. No bill text has been finalized, and differences over accountability provisions have resurfaced.
Congress should abandon any push for federal AI safety mandates that would insert bureaucrats between innovators and the marketplace. Such legislation expands federal power into a domain still taking shape, repeating the pattern of preemptively regulating new technologies before their risks or benefits can be measured by actual use. The Constitution assigns Congress no general authority to police speculative future harms across the entire economy, and the Tenth Amendment leaves most regulatory questions to the states or to private actors. Free markets have driven American technological leadership precisely because they reward results rather than compliance with central directives. Imposing liability rules from Washington before the technology matures will slow progress and concentrate power in agencies that answer to no voter.