DHS Report Highlights AI’s Ability to Amplify Biochemical and Nuclear Threats

The Department of Homeland Security (DHS) recently released a report targeting the mitigation of artificial intelligence's potential to worsen chemical, biological, radiological, and nuclear threats.

Developed with contributions and recommendations from experts in AI and CBRN (Chemical, Biological, Radiological, and Nuclear) issues across DHS, the Department of Energy, private AI labs, academia, and third-party model evaluators, this report focuses on AI-enabled chemical and biological agents. 

It aims to establish longer-term goals for ensuring the safe, secure, and trustworthy development. The report outlines several new measures aimed at AI safety, such as promoting additional credentialing for high-risk scientific databases and establishing standards to address ‘unacceptably dangerous responses’ from large language models.

The recommendations include integrating AI into all phases of planning for countering weapons of mass destruction terrorism and CBRN preparedness, aligning with the National Biodefense Strategy. 

The report also urges federal agencies and commercial providers to responsibly utilize AI in designing, testing, and assessing protective equipment, medical countermeasures (e.g., vaccines, synthetic antibodies).

Furthermore, it advises leveraging DHS’s Artificial Intelligence Safety and Security Board to foster information sharing, establish best practices, and mitigate risks associated with AI technology in CBRN-related dual-use applications that may impact national or economic security.

Earlier this year, DHS also released new guidelines to address AI threats to critical infrastructure and a report on AI misuse in CBRN development.

These guidelines complement the Department's wider initiatives aimed at safeguarding national critical infrastructure and facilitating AI utilization.

Among these efforts is the recent formation of the Artificial Intelligence Safety and Security Board.