On July 26, 2025, China announced its Action Plan for Global Artificial Intelligence (AI) Governance. This initiative builds on Xi Jinping’s October 2023 Global AI Governance Initiative and presents itself as a responsible and benevolent effort to avoid global chaos. However, it reflects China’s broader ambition to shape global AI governance amidst rising technological competition.
At the 2024 Asia-Pacific Economic Cooperation (APEC) summit in Lima, Peru, Chinese leader Xi Jinping and U.S. President Joe Biden agreed that artificial intelligence should never be given the authority to launch nuclear war. However, typical of Biden’s weak foreign policy approach, the agreement was not a formal treaty. Meanwhile, China is actively integrating AI into its defense infrastructures, including its nuclear capabilities.
China has long implemented a “military-civil fusion” strategy to apply AI in military practices, leveraging advancements in civilian AI, where China has become a world leader in areas like facial recognition, speech technology, and deep learning, for military purposes. This approach takes advantage of AI’s dual-use nature and spans various military applications, including command and control, decision-making, and autonomous nuclear weapons systems.
One key focus of China’s AI integration is the development of neural networks for use in hypersonic glide vehicles. Applications being developed in automatic target recognition, autopiloting, missile fusion, and precision guidance. These innovations are expected to enhance the maneuverability and penetration capacity of nuclear missiles.
Additionally, China is exploring how AI can improve missile guidance, target detection, and identification, as well as boost capabilities in electronic warfare and decision-making.
AI technology is enhancing China’s nuclear defensive capabilities by improving situational awareness, early-warning, and missile defense systems. China is particularly concerned about false negatives in early-warning systems, which may fail to detect nuclear attacks due to assumptions about U.S. conventional precision strike capabilities.
In response, China is deploying autonomous interception systems operating at machine speeds to protect its nuclear assets and missile bases from saturation strikes. Additionally, AI-enabled launch-on-warning systems are strengthening the credibility of China’s nuclear retaliation.
China is developing AI systems for command, control, and communications (C3) at three levels. Counter-C3 systems are being designed to enhance cyber capabilities and disrupt opponents’ networks. AI is also being applied at the tactical and operational levels to control uncrewed systems and improve targeting for physical weapons.
China’s nuclear modernization is advancing alongside AI integration, with approximately 350 new ICBM silos expected to be completed by January 2025. As China expands its nuclear arsenal by about 100 warheads per year, these AI-driven advancements are enhancing its ability to penetrate enemy defenses. This expansion, coupled with AI enhancements, is shifting the global nuclear balance from a bipolar U.S.-Russia structure to a tripolar one.
The People’s Liberation Army (PLA) aims to use big data analytics, machine learning, and automation to strengthen the defense of critical networks and amplify offensive cyber operations. Chinese strategists believe well-trained AI programs could unlock new, previously unimaginable methods of cyber-attacks.
At the same time, the United States military is integrating AI into both its defensive and offensive capabilities. U.S. Air Force and Space Force officials see AI as essential in enhancing Nuclear Command, Control, and Communications (NC3) architecture, aiming to speed up decision-making cycles and ensure orders are delivered quickly and securely.
Air Force Gen. Anthony Cotton, head of U.S. Strategic Command, highlighted that AI can process “terabytes of data” to improve decision-making, all while retaining human control.
AI is also being utilized in intelligence and early warning systems for reconnaissance, decision-support, and battlefield data analysis. This includes applications such as coordinating drone swarms and optimizing logistics. However, experts warn that AI systems managing nuclear arsenals could be vulnerable to hacking, potentially leading to unauthorized launches or an inability to respond effectively to real threats.
China is highly adept at hacking, particularly in targeting U.S. defense systems. The Department of Defense’s 2024 annual report on “Military and Security Developments Involving the People’s Republic of China” highlights the significant and persistent cyber-enabled espionage and attack threat China poses to military and critical infrastructure. The report stresses that the PLA views information operations as essential for gaining information superiority early in a conflict, which it sees as crucial for any military campaign’s success.
Another concern is the accelerated decision timelines facilitated by AI, which could leave limited room for diplomatic de-escalation during crises. This speed advantage might fundamentally alter the dynamics of nuclear confrontations.
The concern is not about giving launch codes to AI systems like ChatGPT, but rather that the speed of AI-assisted decisions, combined with potential errors, could amplify the risk of mistakes in critical military decisions.
AI companies are forming partnerships with the U.S. military, with Anthropic announcing classified Claude Gov models for national security in June, and OpenAI collaborating with U.S. national laboratories, including those working on nuclear weapons. China and Russia are also advancing the integration of AI tools into both conventional and nuclear warfare.
These developments continue despite the AI nuclear weapons limitation agreement signed between President Joe Biden and Chinese leader Xi Jinping.
