AI Researchers Must Help Lead Arms Control to Mitigate Military AI Risks

AI researchers have a responsibility to engage in arms control efforts now, not just focus on distant superintelligence risks. The paper argues that while long-term AI safety concerns are important, the immediate challenge of regulating military AI applications demands urgent attention from the research community, as defense contractors are already actively integrating AI into weapons systems.

The defense industry is rapidly forming partnerships between military contractors and AI companies. This "burgeoning coalition" of armament manufacturers and AI firms is accelerating the deployment of military AI systems, creating a near-term governance gap that requires immediate collaborative action between military leaders, diplomacy experts, and AI researchers.

Historical arms control lessons from nuclear deterrence can guide AI military governance. The paper suggests that past successes in reducing catastrophic risks through arms control frameworks—particularly nuclear deterrence—offer valuable models for developing verification methods and diplomatic approaches to manage military AI risks.

Current AI safety solutions are insufficient for military applications. The paper emphasizes that there is a "lack of sufficiently reliable solutions" for ensuring safe military AI deployment, making it critical for researchers to develop new technical approaches to define and mitigate instability in defense contexts.

AI researchers must take a leading role in technical research for military AI verification and security. Rather than leaving arms control entirely to diplomats and military officials, AI experts are uniquely positioned to develop the technical innovations needed to verify compliance and ensure stability in military AI systems.