As competition intensifies over powerful open-weight AI models from China, the Trump administration is relying on a small, fragmented group of officials across different departments to decide on export controls and other restrictions, with no unified policy yet in place. This fragmentation matters because, according to a senior White House official, the disagreement is not a simple two-sided debate but one with "ten sides," meaning that whoever holds the most sway with President Trump personally is likely to end up shaping US AI policy rather than any coordinated interagency process.

Key figures include Commerce Secretary Howard Lutnick, who oversees export controls and has explored incentives for US labs to build open-weight models, while publicly downplaying the capabilities of China's Moonshot AI Kimi K3 model; Arvind Raman, the new acting head of the Center for AI Standards and Innovation, who took over after his predecessor resigned amid efforts to secure Anthropic's Fable 5 model against jailbreaks; and National Cyber Director Sean Cairncross, who takes a harder line on Chinese AI and helped shape Trump's 2 June executive order, with a particular focus on stopping Chinese labs from "distilling" data from top US models. A White House spokesperson said the administration's overarching goal is to widen America's lead in AI innovation.