AI nuclear risk prompts U.S.-China expert safeguard plan

AI nuclear risk has led U.S. and Chinese experts to urge three safeguards before autonomous systems leave leaders minutes to judge an attack.

The recommendations were published last week ahead of planned government-level AI talks and an expected September 24 Washington meeting between President Trump and Chinese President Xi Jinping. The experts want red lines around nuclear systems, human control over consequential cyberattacks and a hotline for autonomous AI incidents.

Melanie Sisson, a senior fellow at the Brookings Institution, and Tianjiao Jiang, an associate professor at Fudan University, took part in a U.S.-China AI and national security dialogue convened since 2019 by Brookings and Tsinghua University's Center for International Security and Strategy. Their proposals move the rivalry beyond chips and models into an arms-control problem: how nuclear-armed rivals identify accidents before treating them as attacks.

Three safeguards before Trump-Xi talks

The package centers on the risk that machine-speed military systems could compress decision time in a crisis. If an AI tool interfered with nuclear command, control and communications, Washington or Beijing might have only minutes to decide whether the other government had ordered the action, the experts said.

The same logic applies to cyber operations against energy, finance and healthcare networks. An autonomous system could malfunction, exceed its instructions or be compromised, while the targeted state might see the activity only as hostile behavior arriving through official channels.

China's State Security Minister Chen Yixin wrote this week in a government outlet that AI "fundamentally transforms the form of military struggle." Chen said AI had shifted from a battlefield support role to a central role, citing AI-directed strikes in the U.S.-Israeli conflict with Iran.

Nuclear red lines widen

Sisson argued that people, not autonomous systems, should retain sole authority to initiate AI-enabled cyberattacks against the other country's nuclear command systems or strategically important infrastructure. Her argument turns on attribution: the receiving side may not be able to tell whether an operation was accidental, unauthorized or deliberate.

Jiang proposed explicit limits barring AI from independently deciding to use nuclear weapons or autonomously attacking nuclear command systems. He also urged both governments to define "meaningful human control" so the same phrase does not conceal different levels of machine autonomy.

The proposal extends a principle endorsed by then-President Joe Biden and Xi in November 2024: humans should control decisions to use nuclear weapons. Lora Saalman, an associate senior fellow at the Stockholm International Peace Research Institute, wrote in February that the point marked a rare U.S.-China convergence even as both sides appeared to be integrating AI more deeply into nuclear-related systems.

AI hotline meets old doubts

Jiang also called for a dedicated military channel for AI incidents. The mechanism would let one government tell the other that unusual automated activity was accidental, unauthorized or still under investigation before it was treated as deliberate state action.

The obstacle is trust in crisis communications already on the books. Carla Freeman at Johns Hopkins School of Advanced International Studies has questioned the existing U.S.-China military hotline, citing Chinese officials' refusal to answer U.S. calls during the February 2023 spy balloon incident.

Freeman wrote in January that China's top-down bureaucracy can leave lower-ranking officers unable to act during encounters with U.S. counterparts. "No one in the Chinese system wants to communicate with the United States until the top leadership in consultation with relevant actors within China's party-state have decided on a response, which may take days if not longer," she said.

Arms-control desks absorb AI

The proposals have not been publicly endorsed by either government. Beijing has placed AI inside the Foreign Ministry's Department of Arms Control, the same bureaucracy that handles nuclear weapons, nonproliferation, missiles and other international security issues.

Shen Jian, China's ambassador for disarmament affairs, told a United Nations meeting in Geneva in June that military AI could affect strategic stability and increase the risk of miscalculation and escalation. China has also argued that AI safety should not become a pretext for technological barriers.

Washington has no equivalent single arms-control home for AI, with policy spread across the White House National Security Council, State Department, Pentagon and other agencies. President Trump has warned that broad restraints could give China an advantage, while Beijing views many U.S. technology controls as attempts to preserve American dominance.

If the two governments agree on limited nuclear red lines, the macro effect would be a narrower channel for crisis escalation rather than a slowdown in AI competition. If talks stall, defense AI suppliers, cloud providers and chipmakers would still face demand from military users, but with fewer shared rules around how their systems behave in a cross-border incident.