When Donald Trump posted on social media last Monday that the United States is "ahead of China and everyone else" in artificial intelligence, he was doing more than boasting. He was signalling a tension that will define the summit between Trump and Chinese President Xi Jinping in Washington next week: how do two rival superpowers manage a technology they each fear the other might misuse, yet neither wants to constrain? AI governance is expected to appear on the agenda, and its inclusion matters. Senior figures in the American technology industry have argued, with increasing urgency, that preventing the worst outcomes from uncontrolled AI systems requires some degree of coordination with Beijing.

Yet the political conditions for a serious agreement are, at best, fragile. Trump has resisted calls for tighter AI regulation, warning that any such effort would hand China a competitive advantage. His position leaves little room for the kind of binding commitments that meaningful governance would demand. The two countries are also developing AI along divergent paths.

American firms are pushing capability frontiers using cutting-edge chips and vast data-centre infrastructure. China, partly because Washington's export restrictions have limited its access to the most advanced semiconductors, has focused instead on cheaper, open-source models designed for wide global adoption — and those models have quietly gained traction among businesses in the United States itself. What, then, can realistically emerge from Washington?