US-China Experts Propose AI Red Lines for Nuclear Systems

Why it matters

As artificial intelligence becomes more autonomous and integrated into military operations, the window for human decision-making during a potential crisis between nuclear-armed powers is shrinking. These expert recommendations represent an early attempt to establish guardrails before AI-driven systems create irreversible escalation scenarios that neither Washington nor Beijing intended.

Security experts from the United States and China are calling for explicit restrictions on artificial intelligence in military systems, warning that autonomous technology could leave governments with only minutes to determine whether they're under attack.

The proposals center on preventing AI from independently controlling nuclear weapons or launching cyberattacks against nuclear command networks. According to experts participating in a US-China dialogue convened by the Brookings Institution and Tsinghua University's Center for International Security and Strategy, an AI malfunction or compromised system could trigger a crisis before humans can assess what actually happened.

Proposed safeguards

Melanie Sisson, a senior fellow at Brookings, argued that humans must retain sole authority to initiate AI-enabled cyberattacks against nuclear command systems or strategically important infrastructure. The risk: an autonomous system could malfunction, exceed its programming, or be compromised, while the targeted country struggles to distinguish between an accident and a deliberate attack.

Tianjiao Jiang, an associate professor at Fudan University, proposed explicit prohibitions on AI independently deciding to use nuclear weapons or autonomously attacking nuclear command systems. He also called for protections around critical infrastructure including energy, finance, and healthcare sectors.

Jiang emphasized the need for a shared definition of "meaningful human control" to prevent both nations from using identical terminology to justify vastly different levels of machine autonomy in military operations.

The speed problem

The experts also recommended establishing a dedicated US-China military hotline specifically for AI incidents. Jiang warned that automated cyber exchanges could escalate faster than humans can intervene—one country's defensive AI might automatically respond to suspicious activity, which the other side could interpret as an attack and retaliate against before understanding the actual situation.

However, the utility of such hotlines remains uncertain. Carla Freeman at Johns Hopkins School of Advanced International Studies has noted that Chinese officials refused to answer US calls during the February 2023 spy balloon incident. She observed that China's rigid bureaucratic hierarchy prevents low-ranking officers from taking initiative, with decisions requiring top leadership consultation that "may take days if not longer."

Divergent approaches

China has increasingly positioned AI within its existing arms-control bureaucracy. The Foreign Ministry's Department of Arms Control now includes AI as a formal portfolio alongside nuclear weapons and non-proliferation. Shen Jian, China's ambassador for disarmament affairs, told a UN meeting in June that military AI could affect strategic stability and raise risks of miscalculation.

Washington has no equivalent centralized arms-control structure for AI, with policy distributed across the National Security Council, State Department, Pentagon, and other agencies.

Both governments remain cautious about measures that could constrain their technological development. Former President Trump warned that broad restraints could give China an advantage, while Beijing views many US technology controls as attempts to preserve American dominance.

China's State Security Minister Chen Yixin wrote this week that AI "fundamentally transforms the form of military struggle," noting it has moved from a supporting battlefield role to a key one.

These details were first reported by Reuters correspondents Eduardo Baptista and Laurie Chen.