U.S. Export Controls Create Paradox for AI Labs' WMD Safety Testing, Experts Urge Clarification

Washington D.C. – U.S. artificial intelligence (AI) laboratories conducting crucial safety evaluations to prevent their models from being used to design or deploy weapons of mass destruction (WMDs) are inadvertently risking violations of U.S. export control laws. This regulatory paradox threatens national security by impeding essential biosecurity testing, according to an analysis by Doni Bloomfield, Joe Khawam, and Tim Schnabel published by Lawfare. They advocate for immediate clarification of existing laws to ensure compliance supports, rather than hinders, safety efforts.

The core issue stems from the broad reach of the Export Administration Regulations (EAR) and the International Traffic in Arms Regulations (ITAR), which govern the transfer of sensitive technologies and information. Many top biosecurity experts are foreign nationals, and sharing information with them—even within U.S. offices—can constitute a "deemed export" requiring government authorization. These licensing processes can take months, creating significant delays for AI labs operating on rapid development cycles.

According to the Lawfare article, "This regulatory paradox threatens national security at a critical moment in AI development." The authors explain that AI biosecurity evaluations involve probing models for capabilities that could assist in biological weapons development, generating artifacts and insights that may fall under export control. Sharing these findings with foreign team members, crucial for comprehensive testing, can trigger licensing requirements from multiple agencies.

The dilemma forces U.S. AI companies to choose between delaying product releases, limiting testing to U.S. citizens and sacrificing evaluation quality, or risking legal penalties. OpenAI, a major AI developer, has reportedly cited export controls as a "regulatory roadblock to safety testing" in a submission to the White House, requesting general authorizations for chemical, biological, radiological, and nuclear risk evaluations.

To resolve this, Bloomfield, Khawam, and Schnabel propose two key steps: issuing clarifying guidance from the Commerce Department’s Bureau of Industry and Security (BIS) and the State Department’s Directorate of Defense Trade Controls (DDTC), and creating targeted regulatory safe harbors. Such measures would confirm that bona fide biosecurity evaluations align with U.S. national security interests and allow approved organizations to conduct essential testing with foreign specialists without undue delays.