January 29, 2024
On January 24, 2024, the James Martin Center for Nonproliferation Studies (CNS) convened a timely seminar to address the intersection of artificial intelligence (AI) and weapons of mass destruction (WMD) nonproliferation.
The half-day event kicked off with introductory remarks by Dr. William Potter, CNS Director, and framing remarks by Senior Research Associate, Dr. Natasha Bajema. She laid out guiding principles for thinking about the relationship between AI and WMD and identified priority research and training areas in which CNS plans to leverage its expertise to help shape how AI tools are developed and used for nonproliferation purposes.
The seminar showcased the wide range of CNS expertise and its collaborative relationships with industry. Dr. Sarah Shoker from Open AI spoke about ongoing efforts to evaluate the potential exploitation of frontier models by nefarious actors seeking WMD capabilities. Dr. Ian Stewart, Head of the CNS DC office, demonstrated how one can leverage cutting-edge AI tools to streamline nonproliferation workflows while Steven de la Fuente discussed how AI approaches can enhance and expedite open-source data collection and imagery analysis. Pivoting to risks, Dr. Allison Berke, Director of the Chemical and Biological Weapons Nonproliferation Program, presented her research assessing the potential dangers of AI-enabled bio-design technologies, which could allow nefarious actors to engineer novel toxins. CNS Scientist-in-Residence Dr. Ferenc Dalnoki-Veress explored classroom applications of AI tools to enhance arms control pedagogy and research through AI agent-based simulations. Dr. Ian Stewart closed out the speaker session by examining emerging policy and governance challenges.
Abbreviated Version of CNS AI research and training agenda in the field of WMD nonproliferation
The James Martin Center for Nonproliferation Studies (CNS) has an ambitious research and training agenda on artificial intelligence (AI) tailored specifically for the field of weapons of mass destruction (WMD) nonproliferation. It identifies priority areas in which CNS aims to leverage its expertise to help shape how AI tools are developed and used for nonproliferation purposes and to support efforts to mitigate new risks for WMD.
The list of initiatives is divided into five key areas:
- Shaping AI Alignment, Safety, and Security, particularly related to chemical, biological, radiological, and nuclear (CBRN) risks. This involves activities like red teaming AI models to test their potential for nefarious use, evaluating AI capabilities for assisting WMD development, and conducting risk assessments of AI tools. CNS has a unique opportunity to leverage its expertise to help shape how AI tools are developed and employed by forming partnerships with industry in areas of mutual interest.
- Leveraging AI and Machine Learning for Nonproliferation Purposes. As more sophisticated AI tools become available, CNS experts will explore ways to use them creatively in the nonproliferation field, including with respect to forecasting proliferation developments, export controls, open-source data mining and content analysis, image recognition, and treaty verification.
- Understanding the Proliferation and National Security Risks of AI. AI is impacting all sectors of society and will increasingly pose significant new WMD proliferation and security risks at home and abroad. CNS will draw upon its exceptional mix of scholars, scientists, and former practitioners to the impact of AI on different facets of CBRN, including both state and non-state actors.
- Informing Governance and Policy Measures: Informed by its hands-on knowledge of AI and deep familiarity with international institutions and diplomacy, CNS will take an active role in conveying to national governments and international organizations what needs to be done to maximize AI’s potential for good while mitigating its negative consequences.
- Educating the Next Generation on the Nexus of AI and WMD. CNS will incorporate new AI tools and their implications for nonproliferation into its academic curricula, training programs, and next-generation engagement. This will equip students, young diplomats, and policy professionals with the knowledge needed to employ AI creatively and responsibly.
Sustaining the CNS AI research and training agenda will require additional funding and the development of new partnerships in industry and the government sector. The challenges posed by AI are enormous as are the opportunities. CNS is committed to building AI awareness, advancing timely policy solutions, and educating students and professionals to effectively address evolving proliferation challenges.
