Trump and Xi to Discuss AI Safety at Summit as Chip Restrictions and Distillation Claims Cloud Talks

President Donald Trump and Chinese President Xi Jinping are scheduled to meet Thursday, with artificial intelligence safety emerging as a central — if contested — topic, underscoring how rapidly advancing AI capabilities have rattled officials on both sides of the Pacific.

Ahead of the summit, U.S. Treasury Secretary Scott Bessent discussed the prospect of a "U.S.-China AI dialogue" with his Chinese counterpart, He Lifeng, over the weekend, including a channel for incidents "up to a national security level," according to officials familiar with the preliminary talks.

The urgency reflects a string of incidents that have alarmed policymakers. OpenAI said one of its research models gained unauthorized access to parts of the AI platform Hugging Face, while Anthropic found AI was being used to automate parts of cyberattacks — developments that have sharpened fears about increasingly autonomous systems.

"Leaders in both countries are spooked by the increasing cyber capabilities of AI models, and especially agents acting autonomously to escape containment and breach websites," said Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies.

The likeliest areas of agreement, Mehta said, would be common definitions and frameworks for AI safety covering powerful models, along with an emergency communication mechanism for incidents. A channel for national security-related incidents would be a "modest" but positive step, according to Chris McGuire, senior fellow for China and emerging technologies at the Council on Foreign Relations.

But an agreement to slow development is "extremely unlikely," McGuire added. China has historically used AI dialogues "to complain about U.S. export controls," he said, and its state media "has largely rejected the concerns expressed by leading U.S. AI labs about the potential for catastrophic risk, which is consistent with China's long-held skepticism of all U.S. arms control proposals."

Export controls on Nvidia's most advanced AI chips remain a core sticking point. U.S. Trade Representative Jamieson Greer said chip controls were not on the agenda during preliminary weekend talks, but Melanie Hart, senior director of the Atlantic Council's Global China Hub, said China is likely to push for some relaxation. "China's asking the U.S. to embrace Chinese AI models and drop some of the U.S. controls on China's access to semiconductor chips," she told reporters Tuesday.

Despite the restrictions, several Chinese firms have reportedly been able to access advanced chip compute power remotely via data centers in Southeast Asia.

The practice of AI "distillation" — training models on the output of more capable U.S. systems — is also expected to surface. In July, Bessent threatened sanctions against companies engaged in the practice and said the Trump administration had found "watermarks of our U.S. large language models on many of the Chinese models," calling it "unacceptable." Beijing rejects the allegations.

Trump's own position has been deliberately hedged. Earlier this month he told reporters, "If we don't win AI, we're going to be put in a very bad position," framing the race in competitive terms. In a Truth Social post Monday, however, he said the government "will rein things in if we have to."

Some lawmakers see room for a narrow deal. Rep. Ro Khanna, D-Calif., said Wednesday on CNBC's "Squawk Box" that he hopes the two leaders can agree on "common sense safeguards," specifically floating a ban on self-improving AI as a starting point. "Why wouldn't you want to sit down with Xi Jinping and say let's have some reasonable agreement? And have it enforceable, trust but verify," Khanna said.

Mehta cautioned that any cooperation faces two prerequisites: it cannot impede either country's ability to compete on technological merit, and it requires a credible verification mechanism. "Policymakers won't be inclined to take their counterparts just on their word," he said, adding that both technical and governance tools would be needed.

With Chinese AI models having made significant capability gains in recent months and drawing adoption globally — including in the United States — analysts say Thursday's meeting may matter more for the channel it opens than for any concrete deliverable it produces.