On October 30, 2023, President Biden signed Executive Order (E.O.) 14110 on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. The overarching goal of the E.O. is “to ensure that America leads the way in seizing the promise and managing the risks of artificial intelligence (AI)” and to establish a governance framework for the safe and responsible development and use of AI.
The Department of Homeland Security (DHS) has played a key role in implementing the E.O. Section 4.4 of the E.O. highlighted the need “to better understand and mitigate the risk of AI being misused to assist in the development or use of chemical, biological, radiological, and nuclear (CBRN) threats – with a particular focus on biological weapons.” Within DHS, the Countering Weapons of Mass Destruction Office (CWMD) is the office responsible for leading DHS efforts and coordinating with domestic and international partners to safeguard the United States against CBRN threats. The President directed DHS, acting through CWMD, to lead the development of an AI CBRN report that evaluated “the potential for AI to be misused to enable the development or production of CBRN threats, while also considering the benefits and application of AI to counter these threats.”
DHS developed the AI CBRN Report through strong collaboration across the U.S. government, academia, and industry. CWMD solicited input from DHS agencies and offices and consulted with experts in AI and CBRN issues from the Office of Science and Technology Policy, Department of Energy, private AI laboratories, academia, think tanks, and third-party model evaluators to evaluate AI model capabilities to present, mitigate, or guard against CBRN threats.
DHS submitted the full report and a fact sheet of selected findings from the AI CBRN report to the President. This report identifies the trends in AI and types of AI models, including foundation models and biological design tools, that might present or intensify biological and chemical threats to the United States. It offers recommendations to mitigate potential threats to national security in the training, deployment, publication, and use of AI models and associated data, as well as underscoring the vital role of safety evaluations and whole-of-community guardrails.