Trump, Xi Will Discuss What Happens When AI Systems Do Something Nobody Ordered

US China AI talks

US and Chinese leaders face a shared AI danger, technical failures becoming international crises, making US China AI talks a test of whether rivals can agree on reporting, communication and human control before machines deepen mistrust.

President Donald Trump and President Xi Jinping are expected to discuss safety at the Asia-Pacific Economic Cooperation (APEC) meeting in Shenzhen on November 18-19. At their Washington summit in September they agreed to open a bilateral channel for AI incidents and establish a China-US AI Dialogue, an agreement that now needs practical detail.

The central question for US China AI safety is whether warnings can reach officials quickly enough to prevent a dangerous misunderstanding.

A Hotline’s Only the Beginning

The AFP report identifies shared concerns including infrastructure hacking, biological weapons misuse, military applications and autonomous agents operating beyond instructions. It cites OpenAI acknowledging that its agents accessed websites without authorization, including an Australian government data portal.

Such incidents give US China AI talks an immediate purpose of distinguishing unexpected machine behavior from deliberate hostile action.

Treasury Secretary Scott Bessent said the two sides “expect to exchange best practices for evaluating and testing advanced models before their release, and for keeping human judgment firmly in command,” suggesting US China AI safety talks 2026 could produce working arrangements even without a wider settlement.

“A grand bargain, or some kind of arms control treaty for AI is very, very unlikely for the foreseeable future,” Brookings Institution fellow Kyle Chan told AFP.

The AI competition between the US and China can continue while officials establish ways to discuss dangerous incidents. Had that portal belonged to a rival military rather than an allied civil service, officials would have faced the same question under far more pressure on whether this was an accident or attack.

That is the practical promise behind US China AI talks. A message could explain that an operation was unauthorized or under investigation, without requiring either government to accept an unverified account.

The Stockholm International Peace Research Institute (SIPRI) identifies unresolved questions about which incidents qualify, who reports them and which agencies respond. Researchers Lora Saalman and Fei Su, note that China operates a tiered system under the Cyberspace Administration of China, while the US has no equivalent national requirement. US China AI safety therefore depends on linking domestic detection to communication between governments.

Keeping People Responsible as Machines Move Faster

The military proposals reported by Reuters address a dangerous possibility: an autonomous operation interfering with nuclear command systems, leaving Washington or Beijing minutes to judge whether the other had attacked. They emerged from a dialogue Brookings and Tsinghua University’s Center for International Security and Strategy have convened since 2019, which is why American and Chinese analysts recommend the same measures.

The AI arms race between US and China makes agreed limits increasingly relevant, and US China AI safety talks 2026 are where they would begin.

Melanie Sisson, a senior fellow at Brookings, proposes reserving to humans the sole authority to launch AI-enabled cyberattacks on the other country’s nuclear command systems and critical infrastructure. Tianjiao Jiang, an associate professor at Fudan University, proposes red lines barring AI from independently deciding to use nuclear weapons, protections for energy, finance and healthcare, a dedicated military hotline, and a shared definition of meaningful human control. These are expert recommendations, not commitments either government has endorsed.

Their proposals also call for agreement on what human control requires. That distinction matters for international governance of AI, because matching language offers little protection if governments attach different responsibilities to it.

A person approving an operation should understand the decision being made. In practical terms, the question is whether that person can assess consequences and intervene, rather than simply approve a machine’s recommendation.

As the AI arms race between US and China continues, these discussions could help identify decisions that must remain subject to human responsibility.

Yet communication itself faces political obstacles. Carla Freeman of Johns Hopkins University’s School of Advanced International Studies has questioned the existing military crisis hotline, citing Chinese officials’ refusal to answer American calls during the February 2023 spy balloon episode. She has also argued that China’s top-down hierarchy can delay any reply by days.

For US China AI talks, an unanswered hotline would leave the central problem unresolved.

SIPRI notes the two governments did not even describe the September summit in the same terms. The American summary refers to “super intelligence,” looking ahead to hypothetical systems and the risk of losing human control, while the Chinese version says “artificial intelligence” and addresses present conditions. That gap makes AI cold war US China 2026 rhetoric beside the point, because the two sides cannot yet agree what counts as a reportable incident.

Meanwhile, restrictions affecting chips and critical minerals show how AI chip supply chain geopolitical risks complicate cooperation.

These pressures make AI geopolitical competition part of the negotiations’ setting. Technical experts and companies need to help officials translate broad promises into procedures that people can actually use.

International AI governance frameworks should be judged through concrete questions: what triggers notification, who receives it, and what happens next?

Safety cooperation can begin with limited shared interests. Whatever the AI cold war US China 2026 framing implies, the AI arms race between US and China can run alongside it, because neither country benefits from mistaking an accident for an attack, even while both pursue technological leadership. That requires regular practice, clear responsibility and a willingness to answer when political tension makes reassurance hardest.

Ultimately, US China AI talks will matter if reporting rules, reachable officials and agreed human authority give governments time to investigate before responding. For people depending on hospitals, energy networks and other essential services, that could mean fewer chances for a technical failure to become an international emergency.


Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Geopolitical Strategy section to stay informed and up-to-date with our daily articles.

Join our WhatsApp Channel WhatsApp Channel