A Department of Homeland Security report released last week said artificial intelligence could unlock secrets to the development of weapons of mass destruction in the wrong hands, particularly chemical and biological threats.
The full report, which was in preparation with the release of a fact sheet in April, examines AI's role in supporting as well as deterring adversary efforts to research, develop and use chemical, biological, radiological and nuclear weapons. The report was requested under President Joe Biden's October 2023 executive order on AI.
“The growing prevalence and capabilities of AI tools could lead to significant changes in the threat landscape to U.S. national security over time, including affecting the means, accessibility, and likelihood of success of CBRN attacks,” the Department of Homeland Security's Countering Weapons of Mass Destruction Office said in the report.
“Known limitations in existing U.S. chemical and biological security regulations and enforcement, combined with the increasing use of AI tools, may increase the likelihood that intentional or unintentional dangerous research outcomes could pose risks to public health, economic security, and national security,” according to the report.
Specifically, the report states that the proliferation of publicly available AI tools could lower the barrier to entry for malicious actors seeking information about the composition, development, and delivery of chemical and biological weapons. While access to laboratory facilities remains a high hurdle, the report notes that so-called “cloud labs” could allow threat actors to remotely develop components for weapons of mass destruction in the physical world under the guise of anonymity.
CWMD recommended that the U.S. “strategically exclude and/or protect sensitive chemical and biological data” from public training materials for large-scale language models and to increase oversight governing access to remotely operated research facilities.
The report also states that specific federal guidance is needed to govern how biological design tools and biologically and chemically specific underlying models are used, which would ideally include “detailed release instructions” for the source code and specifications for the weight calculations used to build the associated language models.
More generally, the report seeks to build consensus within U.S. government regulatory agencies on how to govern AI and machine learning technologies, particularly as they intersect with chemical and biological research.
Other findings included incorporating “safe harbor” vulnerability reporting practices into organizational procedures, conducting internal assessments and red teaming activities, fostering a broader culture of responsibility among the life sciences professional community, and responsibly exploring the potential benefits of AI and machine learning in biological, chemical, and nuclear fields.
The report also envisions a role for AI in mitigating existing CBRN risks through threat detection and response, including disease surveillance, diagnostics, and “many other applications yet to be identified by the national security or public health communities.”
The report's findings are not binding orders, but its contents will help shape future policies and goals within the CWMD office, DHS said.
“Going forward, CWMD will consider how to implement the report's recommendations through existing federal coordination groups and related efforts led by the White House,” a Department of Homeland Security spokesperson said. Next Government/FCW“The department will integrate AI analytics into its existing threat and risk assessments, as well as in planning and procurements it conducts on behalf of federal, state, local, tribal, and territorial partners.”
