Security experts from the United States and China are urging both governments to establish firm safeguards against artificial intelligence triggering an unintended military or nuclear crisis as increasingly autonomous systems enter cyberwarfare, intelligence, command, and weapons networks. The proposals grow out of an ongoing bilateral expert dialogue and would build on the two countries’ existing commitment that humans—not AI—retain control over decisions to use nuclear weapons. Experts are calling for explicit red lines preventing autonomous AI attacks against nuclear command, control, and communications systems; meaningful human authorization for consequential cyberattacks; clearer definitions of human control; and a dedicated U.S.-China communications channel for AI-related military incidents. The central danger is not merely a science-fiction scenario in which an AI independently launches nuclear weapons. A malfunctioning, compromised, unauthorized, or poorly understood AI system could attack or disrupt critical systems, generate misleading threat information, or provoke automated countermeasures so rapidly that national leaders have only minutes to determine whether they are confronting a deliberate enemy attack. Research on the AI-nuclear intersection increasingly warns that integrating AI into military decision-making could compress crisis timelines and amplify miscalculation precisely when deliberate human judgment is most important.
Key Takeaways
- U.S. and Chinese security specialists are proposing explicit prohibitions against AI autonomously initiating attacks on nuclear command systems and are calling for human control over strategically consequential cyber operations.
- The more immediate danger may not be an AI independently deciding to launch nuclear weapons, but AI-generated false information, cyberattacks, malfunctions, unauthorized actions, or automated responses creating the appearance of a deliberate attack and accelerating escalation before leaders can establish what actually happened.
- Washington and Beijing previously agreed that humans should retain control over nuclear-use decisions, but researchers warn that this principle needs practical safeguards—including agreed definitions of meaningful human control, crisis protocols, red lines, and direct communications—before increasingly capable AI becomes deeply embedded in military infrastructure.
In-Depth
Artificial intelligence is rapidly moving from an analytical aid into increasingly consequential military systems, creating a national-security problem that Washington and Beijing cannot safely treat as simply another arena of technological competition. American and Chinese experts involved in a long-running bilateral dialogue are advocating rules designed to prevent AI from inadvertently producing a military confrontation that human leaders never intended.
Their proposals begin with an important principle: machines should not possess autonomous authority over nuclear weapons or strategically consequential attacks against nuclear command-and-control infrastructure. But the problem extends beyond the proverbial nuclear button. AI could penetrate intelligence analysis, cyber operations, early-warning networks and defensive systems. An autonomous cyber operation could malfunction, exceed its instructions or be manipulated by a third party. The targeted nation might then interpret the resulting disruption as an intentional American or Chinese attack.
That possibility becomes particularly dangerous because AI operates at machine speed. Two automated systems responding to one another could escalate a confrontation in milliseconds, while political and military leaders require considerably longer to determine attribution and intent.
The United States and China therefore have a legitimate shared interest in establishing enforceable boundaries without surrendering their ability to develop defensive AI capabilities. A dedicated military AI hotline, agreed red lines surrounding nuclear infrastructure and genuine human authorization requirements could provide basic guardrails.
The objective should not be broad international bureaucracy controlling American technological development. It should be narrower and more practical: preserve human responsibility for decisions capable of starting wars, maintain reliable communication between nuclear powers and prevent an algorithmic mistake from becoming an irreversible strategic catastrophe.
Sources
- https://www.brookings.edu/articles/advancing-human-control-of-military-ai/
- https://www.sipri.org/publications/2026/other-publications/ai-chinese-indian-and-us-nuclear-postures-norms-and-systems
- https://www.cnas.org/publications/reports/averting-ai-armageddon
- https://www.armscontrol.org/act/2025-09/features/artificial-intelligence-and-nuclear-command-and-control-its-even-more