Rep. Celeste Maloy Introduces the ATOMIC Act to Strengthen AI Nuclear Security

This week, Rep. Celeste Maloy (UT-02) introduced the bi-partisan AI Threat Output and Monitoring Incident Containment Act (ATOMIC Act), legislation directing the Department of Energy to establish a national program to evaluate the most advanced artificial intelligence systems for nuclear-related risks before those risks become national security incidents. The bill is being co-led by Rep. Sara Jacobs (CA-51).

The ATOMIC Act creates the Advanced Artificial Intelligence Nuclear Evaluation Program, which would leverage the Department of Energy's National Laboratories to test advanced AI systems for vulnerabilities such as generating sensitive nuclear information, bypassing safety guardrails, exhibiting deceptive behavior, or acting unpredictably in high-stakes environments. The program would also help develop practical containment protocols, mitigation strategies, and technical standards to address emerging AI-related nuclear threats.

"As AI capabilities continue to advance, we have a responsibility to understand the risks before they become real-world crises," said Rep. Maloy. "The United States already has the world's leading nuclear security experts in our National Laboratories. This bill builds on that expertise to ensure we can identify emerging threats, strengthen safeguards, and have the information we need to make informed decisions about future AI policy."

The legislation applies only to the largest developers of advanced AI systems, specifically those that have invested at least $2 billion in AI development over the previous five years. Covered developers would provide the secure access and technical information necessary for meaningful evaluations, while the bill includes strong protections to safeguard proprietary business information and trade secrets.

"No one wants another Three-Mile Island or any other major nuclear incident," said Rep. Jacobs. "As AI becomes more powerful, we have a responsibility to identify and guard against new risks before they become real-world threats. That's why I'm proud to co-lead the bipartisan ATOMIC Act to identify AI-related nuclear risks, establish safeguards, and develop mitigation plans. These common-sense guardrails are an important first step to help us harness AI for good while protecting against worst-case scenarios."

Under the bill, the Department of Energy would provide Congress with an initial report within one year identifying trends, capabilities, and risks observed through the evaluation program, along with recommendations on whether additional legislation is warranted. Annual updates would continue throughout the program's seven-year duration.

The ATOMIC Act has been endorsed by the AI Policy Network.