Trump, Xi and the fight over who controls AI

Two days before welcoming Chinese President Xi Jinping to the White House, Donald Trump addressed the UN General Assembly and rejected a globalist scheme controlling artificial intelligence (AI). He dubbed the technology "superintelligence" and touted an oil deal with Venezuela following a military operation that removed Venezuela's president, Nicolas Maduro. Trump dismissed any warnings of catastrophe, dismissing them as the same voices who predicted the doom of global warming.

This prelude to the upcoming summit between Trump and Xi Jinping reveals the stark contrast in approaches to AI. Trump has refused to accept shared rules for AI, while China has signed agreements with 28 nations to establish the World Artificial Intelligence Cooperation Organisation (Waico). Washington is holding the reins through export controls, paid access to strong models, and the notion that US systems should only reach trusted partners.

China, in response, argues that each new restriction only accelerates domestic capability. Beijing offers cheap or free models to attract the global south through BRICS and a UN governance framework. This approach contrasts with Washington's paid ladder of access to AI models.

The geopolitical stakes are high, particularly with China controlling 70% of rare earth mining, 85% of refining, and nearly 90% of magnet production. These materials are crucial to semiconductor chips and the defense sector. Beijing has tightened its grip on these materials, causing magnet shipments to the US to fall by 21% in August. The squeeze is timed with a tariff truce expiring on November 10, and a Pentagon rule barring rare-earth magnets, tungsten, tantalum, and molybdenum from January 1, 2027.

On AI safety, the meeting between Trump and Xi Jinping is expected to focus on managing frontier AI risk and protecting intellectual property (IP). The leaders are skeptical of AI-related catastrophes, but the summit's potential achievements are narrow. They may agree on a notification mechanism for AI incidents reaching national security levels and reaffirm the 2024 consensus on keeping AI out of nuclear command.

However, the broader contest of visions between the US and China is stark. While Washington seeks a private channel with Xi, it has rejected a longer moratorium on rare-earth exports and a ban on Chinese minerals in defense supply chains. China, on the other hand, offers an open door to AI, distributing models that could set de facto standards.

Observers worry that energy demand, lithography, advanced packaging, researcher flows, and controlled inputs like gallium and germanium may be overlooked in the summit's discussions. Neither side seems keen to raise these concerns, leading to a wider contest of visions. The June G7 summit in Evian leaned towards a club of access for reliable democracies, while the September BRICS summit in New Delhi emphasized sovereignty, equity, and south-to-south cooperation.

India, as BRICS chair and G7 partner, sits at the intersection of these competing visions. It hosted its own AI Impact Summit in February, advocating for a body that gives emerging markets a real seat in rule-making. India's concerns include reliable compute access, exposure to mineral pressures, and the fear that a two-rule-makers deal on standards would leave everyone else as rule-takers.

The hope is that the channel opened between Washington and Beijing becomes a floor on which others can build, rather than a perimeter wall around a private club.