Experts Urge US, China to Ban AI Nuclear Weapons Control

Brookings Institution researchers from both the United States and China have issued a joint call for Washington and Beijing to establish a mutual agreement ensuring that artificial intelligence will never be granted autonomous control over nuclear weapons. This initiative, detailed in a recent policy brief, aims to prevent a dangerous escalation in the development and deployment of AI-powered military systems, particularly those with the potential to trigger catastrophic conflict. The core of the proposal is a commitment to maintaining meaningful human control over the decision-making processes for nuclear arsenals, thereby mitigating the risks associated with AI's increasing integration into strategic defense systems.

The researchers highlight that as AI capabilities advance, the temptation to delegate critical, time-sensitive decisions to automated systems in a high-stakes environment like nuclear command and control could become significant. However, they argue that the inherent unpredictability and potential for emergent behavior in complex AI systems make them unsuitable for such ultimate authority. The potential for algorithmic errors, unforeseen interactions between systems, or even deliberate manipulation by adversaries could lead to accidental or unauthorized nuclear launches, with devastating global consequences. The proposal specifically targets the two largest nuclear powers, recognizing their pivotal role in shaping global security norms and arms control frameworks.

This call for a bilateral agreement is situated within a broader international discussion about the ethical and security implications of artificial intelligence in warfare. While AI offers potential benefits in areas such as intelligence analysis, logistics, and even defensive systems, its application in offensive or command-and-control roles, especially concerning weapons of mass destruction, raises profound concerns. The Brookings researchers emphasize that a proactive, cooperative approach between the US and China is essential to establish a precedent and prevent a future arms race in AI-driven nuclear capabilities. They suggest that such an agreement could serve as a foundation for broader international dialogue and potentially lead to multilateral treaties on the responsible development and use of AI in military contexts.

The policy brief underscores the urgency of addressing these issues, noting that the pace of AI development is rapid and the window for establishing robust safeguards may be closing. By advocating for a clear, verifiable commitment to human control, the researchers aim to foster trust and transparency between the two nuclear superpowers, reducing the likelihood of miscalculation and enhancing global stability. The proposal is not intended to halt AI research or development but rather to channel it away from the most perilous applications, ensuring that human judgment remains paramount in decisions of existential importance.