US and China Experts Propose Nuclear-Style AI Safety Safeguards

Red Lines Around Nuclear Systems

Humans should retain sole authority to initiate AI-enabled cyberattacks against opposing nuclear command, control, and communications systems or strategically important infrastructure, according to Melanie Sisson, a senior fellow at the Brookings Institution. An autonomous system could malfunction, exceed its instructions, or be compromised, leaving the targeted nation struggling to determine whether an action was accidental or ordered by a rival.

Fudan University associate professor Tianjiao Jiang suggested establishing clear red lines to ensure AI does not autonomously decide to launch nuclear weapons or target nuclear command systems. These protections would extend to critical infrastructure, including energy, finance, and healthcare. Jiang also urged both countries to establish a shared definition of "meaningful human control" to prevent Washington and Beijing from using identical language to justify varying levels of machine autonomy.

These recommendations build on a principle endorsed by then-U.S. President Joe Biden and Xi in November 2024, agreeing that humans should retain control over decisions to use nuclear weapons. In February, Lora Saalman, an associate senior fellow at the Stockholm International Peace Research Institute, noted that this agreement marks a significant area of alignment between the U.S. and China, even as AI is more deeply incorporated into the nuclear-related systems of both countries.

Managing Machine Speed and Communication Risks

Jiang further recommended a specific U.S.-China military hotline for AI-related incidents, cautioning that human intervention might be too slow to stop the rapid escalation of automated cyber exchanges. Under this scenario, one country’s defensive AI might automatically respond to suspicious activity, prompting the other side to interpret the response as an attack and retaliate before understanding the situation. A hotline would allow one government to notify the other that an unusual AI operation was accidental, unauthorized, or still under investigation.

However, experts have questioned the utility of existing crisis communication channels. Carla Freeman at the Johns Hopkins School of Advanced International Studies noted that Chinese officials refused to answer U.S. counterparts’ calls regarding the "spy balloon incident" in February 2023. Writing in War on the Rocks in January, Freeman pointed out that China’s rigid top-down hierarchy often disempowers low-ranking officers from taking initiative during engagements 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," Freeman wrote.

Divergent Bureaucratic Approaches in Washington and Beijing

Neither government has publicly endorsed the Brookings-Tsinghua proposals, and both remain wary of measures that could slow their own technological development. While Trump has cautioned that extensive restrictions could give China an edge, Beijing views many of the technology controls imposed by the U.S. as efforts to maintain American hegemony.

China has increasingly integrated AI into its formal arms-control bureaucracy, handling the technology through the Foreign Ministry’s Department of Arms Control alongside nuclear weapons, non-proliferation, and missiles. 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 raise the risk of miscalculation, adding that weapons should remain under human control while warning that AI safety should not become a pretext for technological barriers.

By contrast, Washington has no single arms-control home for AI. National security policy remains dispersed across the White House National Security Council, the State Department, the Pentagon, and other federal agencies. Meanwhile, China’s State Security Minister Chen Yixin wrote in a government outlet that AI fundamentally transforms military struggle, citing AI-directed strikes in the U.S.-Israeli war with Iran as proof that the technology has moved from a supporting battlefield role to a key operational function.