US, China security experts propose nuclear-style safeguards for AI risks

BEIJING, Sept 17 - An AI system that interferes with a nuclear command network or launches a military cyber operation could leave Washington or Beijing with minutes to decide whether the other government attacked, according to U.S. and Chinese security experts.The experts, part of a U.S.-China dialogue whose work helps inform official discussions, are urging both governments to prepare for such scenarios amid rising concerns about increasingly powerful and more autonomous systems.Their proposals include red lines around nuclear systems, human control over consequential cyberattacks and a hotline for incidents involving autonomous AI.The recommendations were published last week ahead of planned government-level AI talks and an expected September 24 meeting in Washington between U.S. President Donald Trump and Chinese President Xi Jinping.Authors Melanie Sisson, a senior fellow at the Brookings Institution think tank, and Tianjiao Jiang, an associate professor at Fudan University, participated in the 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 show how the U.S.-China AI rivalry is moving beyond competition over chips and advanced models into questions long associated with arms control: how to prevent accidents from being mistaken for attacks, limit machine-made military decisions and set communication protocols for nuclear-armed rivals when technology behaves unexpectedly.China's State Security Minister Chen Yixin this week wrote in a government outlet that AI "fundamentally transforms the form of military struggle". He said it had moved from playing a supporting role on the battlefield to a key role, citing AI-directed strikes in the U.S.-Israeli war with Iran.RED LINES AROUND NUCLEAR SYSTEMSBrookings Institution's Sisson argued that humans should retain sole authority to initiate AI-enabled cyberattacks against the other country's nuclear command, control and communications systems or strategically important infrastructure.An autonomous system could malfunction, exceed its instructions or be compromised, while the country under attack might struggle to tell whether the action was accidental or ordered by its rival.Fudan University's Jiang proposed explicit red lines preventing AI from independently deciding to use nuclear weapons or autonomously attacking nuclear command systems, along with protections around critical infrastructure including energy, finance and healthcare.He also urged a shared definition of "meaningful human control" to prevent both countries from using identical language to justify different levels of machine autonomy.The idea builds on a principle endorsed by then-U.S. President Joe Biden and Xi in November 2024 that humans should retain control over decisions to use nuclear weapons.Lora Saalman, an associate senior fellow at the Stockholm International Peace Research Institute, wrote in February that this was an important point of U.S.-China convergence even as both countries showed signs of integrating AI more deeply into nuclear and nuclear-related systems.WHEN MACHINES MOVE FASTER THAN HUMANSJiang also proposed a dedicated U.S.-China military hotline for AI incidents, warning that automated cyber exchanges could escalate too quickly for human intervention.One country's defensive AI might automatically respond to suspicious activity, while the other side could interpret that response as an attack and retaliate before understanding what actually happened.A hotline could allow one government to tell the other that an unusual AI operation was accidental, unauthorised or still under investigation before it was treated as deliberate state action.However, experts including Carla Freeman at Johns Hopkins School of Advanced International Studies have questioned the utility of the existing U.S.-China military crisis communication hotline, citing Chinese officials' refusal to answer U.S. counterparts' calls about the "spy balloon incident" in February 2023.The Chinese bureaucracy's rigid top-down hierarchy also disempowers low-ranking officers from taking the initiative during engagements with U.S. counterparts, she wrote in an article for War on the Rocks in January."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.VIEWS IN WASHINGTON AND BEIJINGNeither government has publicly endorsed the Brookings-Tsinghua proposals, but China has increasingly placed AI inside its existing arms-control bureaucracy.AI is a formal portfolio of the Foreign Ministry's Department of Arms Control, which also handles nuclear weapons, non-proliferation, missiles and other international security issues.Shen Jian, China's ambassador for disarmament affairs, told a U.N. meeting in Geneva in June that military AI could affect strategic stability and raise risks of miscalculation and conflict escalation, and that weapons should remain under human control. China also argues that AI safety should not become a pretext for technological barriers.Washington has no equivalent single arms-control home for AI, with national security policy spread across the White House National Security Council, State Department, Pentagon and other agencies.Both governments remain wary of measures that could slow their own technological development. Trump has warned broad restraints could hand China an advantage, while Beijing views many U.S. technology controls as attempts to preserve American dominance. REUTERS