Trade truce extended to January 10 gave room for AI talks between Trump and Xi

Trade truce extended to January 10 gave room for AI talks between Trump and Xi

Donald Trump and Xi Jinping met at the White House on September 24, 2026, with artificial intelligence at the top of the agenda – even as multilateral initiatives for frontier AI governance, including a Norwegian-Finnish call to action, stood outside the negotiating room. The question is whether the most important AI agreements going forward will be struck by two powers alone.

The Meeting That Is Confirmed

Here is what has been established: President Donald Trump and China's leader Xi Jinping met at the White House on Thursday, September 24, 2026, for talks that according to CNBC's coverage concerned trade, artificial intelligence, and the Iran war. Trump himself confirmed on Truth Social that AI would top the agenda. The opening ceremony and leaders' remarks were documented by AP's live coverage from Washington.

Before the summit itself, the framework had already been set. Treasury Secretary Scott Bessent said on Wednesday evening on Fox News, as reported by CNBC, that the two countries had agreed to extend their temporary trade truce by two months, to January 10. The truce, which had lowered US tariffs and suspended Beijing's export controls on rare earths and critical minerals, was originally set to expire in mid-November. The extension gave the negotiations room to maneuver – and simultaneously made the summit a power play in which both sides had something to lose from a breakdown.

The Chinese official readout, as relayed by AP, said in essence that Xi stated the two leaders had decided to build a "constructive, strategic and stable" relationship between China and the United States – a relationship that, according to the statement, benefits not only the Chinese and American peoples, but the people of the world. The phrasing is worth noting: the very words "the people of the world" appeared in a document from a meeting at which the rest of the world was not represented.

What the AI Talks Are Said to Cover

The details of the AI portion of the agenda come mainly from a single source: an opinion column syndicated on MSN titled "US-China AI talks: A two-power bargain that could leave the world with no say." That means much of what follows must be read with the author's analysis as the starting point – not as confirmed fact.

According to the column, managing frontier AI risk and protecting intellectual property top the agenda. The telling detail, the columnist writes, is "who asked for" the meeting: Chinese state media claim the US pushed for the talks because Washington fears Beijing is gaining the upper hand. This account rests solely on Chinese state media and has not been confirmed from the American side.

More concretely: On the Sunday before the summit, Bessent met with China's Vice Premier He Lifeng. According to the column, a notification mechanism for AI incidents reaching national-security level was floated there – an arrangement under which the countries would inform one another if an AI incident had consequences beyond the commercial. Such a mechanism would, if so, have been one of the first operational agreements between great powers on frontier AI risk. But it is not confirmed in any official communiqué in the available source material.

The columnist also writes that a statement reaffirming the 2024 consensus on keeping AI out of nuclear command chains is "plausible" – an explicitly speculative statement, not an announced deliverable from either government. It is worth stressing: Nothing in the documented source material confirms concrete AI deliverables from the summit itself. What exists is the agenda (via Trump's own Truth Social post), the background negotiations, and the column's retelling.

The Bargaining Chips: Rare Earths and Tariff Deadlines

The reason AI landed on the summit agenda between two trading partners is that technology issues and trade policy have become intertwined. The truce extended to January 10 was precisely about rare earths – materials critical to semiconductors, and thus to the entire AI infrastructure.

The column argues that China controls roughly 70 percent of rare earth mining, 85 percent of refining, and nearly 90 percent of magnet production – described as an iron grip on materials vital to the chip industry and the defense sector. These figures, however, are not tied to a named source in the column and are not corroborated by other sources in the available material. They should be read as the columnist's estimates, not as documented market data.

A chronological ambiguity also arises here that should be laid openly on the table: CNBC, citing Bessent, writes that the truce was extended by two months to January 10, while the column refers to it expiring on November 10. The most likely explanation is that the two sources describe the situation before and after the extension – but this is not reconciled in the documents, and readers should know the discrepancy exists.

It is this interweaving that makes the AI question bilateral: market access to chips, minerals, and models are bargaining chips. In the same week, Chinese authorities opened a probe into DeepSeek and Moonshot AI, according to the column, after Anthropic accused the two Chinese model developers of routing sensitive data through the company's Claude models. Anthropic's accusation is a company statement, and the probe is known only through the column's retelling – but it illustrates how commercial AI disputes are now treated as security policy.

The Counterpoint: Norway and Finland Outside the Room

The same week Trump and Xi negotiated bilaterally, Norway's Prime Minister Jonas Gahr Støre and Finland's President Alexander Stubb launched an initiative called "A Call for Control of Frontier AI Models." According to the column, the call is endorsed by leaders from Germany, Canada, Singapore, the Netherlands, and South Africa, in addition to the president of the European Commission and others. It urges shared standards, transparency about serious security incidents, and an international body to verify limits when capability thresholds are crossed.

For Norwegian readers, the tension is direct: Støre was thus working, in the same week, for shared standards and international verification, while the two countries that actually hold the frontier models and the computing capacity negotiated at the White House with no others present. Neither party in the bilateral game is a signatory to the Nordic-led call.

China has its own multilateral structure. According to the same column, China signed an AI partnership with 28 mostly developing countries in Shanghai in July, to establish the World Artificial Intelligence Cooperation Organisation (Waico). Waico represents an alternative governance foundation – but it in turn excludes Western frontier developers, and the talks in Washington took place without reference to it.

Trump, for his part, has rejected multilateral AI governance. Two days before the summit, on September 22, he stood – according to the column's rendering – before the UN General Assembly and dismissed what he called a "globalist scheme to control [AI]," renaming the technology "superintelligence" in official American terminology. The quote rests on a secondary rendering in the column – no speech transcripts are available in the source material – but the substance, the rejection of shared rules two days before a bilateral summit, is the column's core observation and deserves to be taken seriously as context.

Open Questions

What is not known is nearly as important as what is. There is no government communiqué on concrete AI outcomes from the summit in the available material. Whether the notification mechanism from the Bessent–He Lifeng meeting was formalized is unknown. Whether the nuclear command statement was actually issued is the columnist's speculation. And the account that the US requested the talks rests exclusively on Chinese state media.

Nor is there any documentation in the material that either party at the White House took the multilateral track into account – neither the Nordic-led call, the UN track, nor Waico. The column's thesis is that a two-power game can leave the rest of the world without real influence over frontier AI governance, and its title says so deliberately: A two-power bargain that could leave the world with no say. The thesis is plausible given the chronology, but it is analysis, not documented fact.

For Norway, the consequence is concrete and uncomfortable. The country has aligned itself with the track for shared standards and international verification, with support from leaders across continents. But as of this writing there is no indication that the track has any entry point into the process actually shaping the framework: negotiations between Washington and Beijing, in which technology, minerals, and market access are treated as bargaining chips. After the September 24 summit, the most important decisions on frontier AI governance remain – at least for now – a matter for two.