Trump, Xi Race to Build AI Incident Hotline Before Shenzhen

By Daniel Rao | October 16, 2026

Donald Trump does not want to "work with" Xi Jinping on governing AI. He said so on the record this month: America is "leading by at least a year, maybe a year and a half," and when you're leading, "you don't open it up to each other." Which is what makes this morning's development so deliciously contradictory: his technical teams are in regular, working-level talks with Beijing to build an AI incident hotline — and both sides want to unveil it as a deliverable when the two presidents meet on the sidelines of the APEC summit in Shenzhen on November 18–19.

The exclusive, published today by the South China Morning Post and attributed to three anonymous sources, says AI has "taken on a more prominent role" in the bilateral agenda. A framework for handling AI-related incidents — the backchannel that would activate when an AI system does something alarming across borders — plus a standing bilateral dialogue could be presented as the tangible outcome of the leaders' third face-to-face meeting this year.

Think about that for a second. The same administration that renamed artificial intelligence "super intelligence" because "artificial" sounds "fake," put a sitting intelligence chief in charge of AI policy, and ordered every AI company to "immediately disclose" model incidents as a "critical national security obligation" is now negotiating the fine print of cooperation with its main rival in the race.

An AI hotline as the Shenzhen deliverable

The Post's sources say technical teams from both countries are in regular contact to iron out the details of a framework announced last month for handling AI-related incidents and institutionalising bilateral dialogue. The goal, one source said, is to have a basic framework for responding to AI-related crises in place before the Shenzhen meeting, with the operational details finalised afterward.

Another possible outcome, according to the Post: a regular AI cooperation mechanism involving government officials, research institutions and business representatives. The exchanges at the expert level, three sources said, intensified sharply after Xi Jinping's state visit to Washington in September — and AI has been climbing the agenda ever since.

Photo: Xi Jinping and Donald Trump during Xi's September 2026 state visit to Washington. Official White House photo by Margo Martin, public domain.

Beijing, Washington, Shenzhen: three handshakes in eight months

The Shenzhen meeting would be the third face-to-face encounter between the two leaders this year. In May, Trump made a state visit to China, appearing with Xi at a welcome ceremony in Beijing's Great Hall of the People. In September, Xi returned the visit: a state visit to Washington — the first by a Chinese leader in 11 years — with a September 24 arrival ceremony on the White House South Lawn and a state dinner in the East Room.

The official White House photograph of that visit holds a telling detail: a model jumbo jet on the coffee table between the two leaders, arranged on facing armchairs like rival CEOs about to negotiate. The pageantry has been relentless on both sides. After Shenzhen, the calendar continues: the two men are also expected to meet at the G20 at the Trump National Doral resort in Miami in December, with the current trade truce running until January 10, 2027.

What September's eight-point deal actually promised

We checked the Chinese-language eight-point account of that September summit (via the Asia Society's translation) against the White House fact sheet, and this time the two sides are describing the same deal in different vocabulary: both versions list a bilateral dialogue on AI risks and benefits and a dedicated communication channel for AI-related incidents, with the next exchange due by November. For a relationship where the two governments routinely narrate different summits, that alignment is itself news.

The White House version, dated September 25, called it the "U.S.-China Super Intelligence Dialogue" — note the rebranded terminology — and said the next exchange would occur by November 2026. Beijing's account calls it the "China-U.S. AI Dialogue," with the next round in November and an incident communication channel attached. The Post notes, honestly, that these initiatives "remain largely declarative for now." That is the gap the October working-level talks are trying to close.

When the Chinese embassy in Washington was asked about the talks on Thursday, it told the Post that Beijing "remains open to maintaining communication" through the dialogue mechanism — and added, pointedly, that "China puts equal emphasis on development and security in terms of AI. We take seriously the inherent and secondary risks of AI." As of this morning, no fresh White House or State Department comment on the talks has surfaced publicly.

Trump's awkward contradiction: no joint governance, but an incident hotline

This is where the story gets genuinely strange. On the record, Trump has been emphatic that cooperation with Beijing on AI is off the table. He does not want to work with Xi on governing AI technology, Reuters reported this month. And during the September summit week he put it even more bluntly, according to an Associated Press account: "They want to stop our progress because we're leading China by a lot and we're going to keep it that way." He added: "I would rather not integrate because we're leading by a lot. When you're leading, you don't open it up to each other. We're leading by at least a year, maybe a year and a half. Some people say two years. I'm not looking to open it up."

Yet the channel now taking shape was, in a sense, a US idea first. Treasury Secretary Scott Bessent floated a bilateral AI communication channel to Vice Premier He Lifeng in New York on September 20 — four days before the summit. This blog covered that proposal at the time. The September deal turned the proposal into a bilateral commitment. The October talks are the unglamorous next step: turning a commitment into machinery.

The distance between "I'm not looking to open it up" and negotiating an incident hotline is the distance between campaign-trail posture and the quiet logic of crisis management. A hotline does not end rivalry; it manages it. The model here is not an alliance — it is the US–Soviet nuclear hotline, which existed precisely because the two sides distrusted each other profoundly. Xi himself framed the rivalry at the summit as something to be steered: the two powers, as "leading nations in artificial intelligence," share "the capability and responsibility to develop and manage AI for good," he said, according to Al Jazeera's reporting, calling for "a healthy competition" that acts as "a race of catching up with one another, not a wrestle in which one either wins or loses."

Clayton's Super Intelligence Force, in the room

There is a second, less noticed thread converging on Shenzhen. Earlier this month, Trump named Director of National Intelligence Jay Clayton as his AI czar, to lead a newly created "Super Intelligence Force" — the panel tasked with reporting within 120 days on AI risks and opportunities while, as Clayton put it to the Wall Street Journal, ensuring that "the risk of not being first is high." We covered the appointment here.

The Force sits at the intersection of everything this story touches. Its charter includes reviewing federal reporting mechanisms for AI-related breaches, hacks and other incidents — the domestic side of the same incident problem the Shenzhen channel would handle internationally. And the White House has separately ordered American AI companies to "immediately disclose" model incidents as a "critical national security obligation" — a directive we reported in detail earlier this month.

Photo: Shenzhen at night. The APEC Economic Leaders' Meeting takes place in Shenzhen on November 18–19, 2026. CC BY 3.0.

So the picture forming is this: mandatory incident reporting at home, an incident hotline with Beijing abroad, and a sitting intelligence chief running point on both. It is a coherent architecture — if you believe the cooperation half will survive contact with the rivalry half.

Our take: can a hotline survive the tech war?

Here is the honest question, and it is the one that will decide whether Shenzhen produces crisis infrastructure or summit decoration: can an AI incident hotline function inside a relationship that is otherwise defined by export controls, chip bans, a simmering tariff war and mutual espionage charges?

The precedents are mixed, and mostly unflattering. Cold War hotlines worked because both sides, however bitterly, shared one overriding fear. The US and China share something similar now — both are terrified of an AI incident neither can explain or contain. But the Post's own reporting concedes the foundation is shoddy: the initiatives are "largely declarative," the operational details don't exist yet, and everything the paper knows comes from anonymous sources talking about talks that neither government has publicly confirmed in any detail.

There is also the very real possibility that this is process theatre ahead of Shenzhen — a deliverable manufactured because summits demand deliverables. Wang Zichen, deputy secretary-general of the Beijing-based Center for China and Globalization, told the Associated Press that the emphasis on "working-level engagement, military-to-military dialogue and crisis-management mechanisms suggests an effort to make that stability more durable and institutionalized." That is the generous reading. The ungenerous one is that both sides needed something to announce in Shenzhen that isn't about tariffs.

My read: the hotline is worth having even if born as theatre. The US–Soviet hotline was also, at first, a piece of summit choreography. It became indispensable later, when the crisis it was built for finally arrived. The honest test of the Shenzhen announcement won't be the communique — it will be whether the first real AI incident gets reported through the channel instead of buried by it. If it does, this will have been one of the most quietly important agreements of the year. If it doesn't, it was decoration.

The questions everyone is asking

When exactly is the Shenzhen meeting?

The APEC Economic Leaders' Meeting takes place in Shenzhen on November 18–19, 2026. Trump and Xi are expected to meet on the sidelines — their third in-person meeting this year, after Beijing in May and Washington in September. The two are also expected to meet again at the G20 at the Trump National Doral resort in Miami in December.

What would an AI incident hotline actually do?

No public technical spec exists yet, which is part of the point of the current talks. The concept, as announced in September, is a bilateral communication channel for AI-related incidents plus a standing dialogue on AI risks and benefits. One source told the Post the goal is a basic framework for responding to AI crises before Shenzhen, with operational details finalised afterward. Think incident-notification line, not joint development programme.

Is the deal already done?

No. The South China Morning Post itself stresses the initiatives "remain largely declarative for now," and everything about the current talks rests on anonymous sources. The deliverable is aimed for, not secured. Watch whether both governments publicly confirm the framework between now and November 18.

Why does Washington keep saying "super intelligence" instead of AI?

Trump's rebrand. He has called "artificial intelligence" an "unfortunate and inaccurate" term — "artificial" makes intelligence sound "fake" — and pushed for "super intelligence" in all US documents. The September White House fact sheet duly calls the channel the "U.S.-China Super Intelligence Dialogue"; Beijing's account sticks with "AI Dialogue." Same deal, different vocabulary.

Sources

Reporting in this article relies on the South China Morning Post's October 16 exclusive (three anonymous sources); the White House September 25 fact sheet as reported by IANS/Yonhap; the Chinese government's eight-point account as translated by the Asia Society's Center for China Analysis; the Associated Press's September 26 summit coverage; Al Jazeera's summit reporting; and Reuters' October 3–4 coverage of the Super Intelligence Force appointment. Where facts rest only on anonymous sources, we have said so.

---

Daniel Rao covers the US–China technology rivalry and the politics of AI for TechPulse USA.