Why Trump and Anthropic Became Unlikely Allies on Open-Source AI

You may have seen the news this week: the United States is moving to restrict open-source AI models. What made people raise an eyebrow was a separate detail: Donald Trump and Anthropic, rarely on the same page, ended up in the same camp. When an AI model releases its weights publicly, it's the equivalent of publishing the recipe. Anyone can download and run it, and governments lose the ability to control who uses it or how. Chinese companies can download, fine-tune, and distill these models, releasing competing versions at lower cost. Scott Galloway put it in historical terms: "What used to take 20 years now takes two months." His example: Chinese companies obtained the intellectual property for Siemens base stations, manufactured them at lower cost, and sold them back to global telecoms at 40 cents on the dollar, capturing the market. AI is following the same playbook, but with two decades of time compressed out. His read on China's strategy: they don't need to build a better AI. They just need to make American AI worthless. Open-weight models like DeepSeek claim 90% of leading US model performance at a third of the price. If enough large enterprises cancel contracts with OpenAI and Anthropic, the valuations collapse. Roughly 40% of the S&P 500 is tied, directly or indirectly, to the American AI bet. Pull that thread, and you get a recession. Blocking open source is the US government's way of keeping the gate. Anthropic genuinely believes the most dangerous AI capabilities shouldn't be freely available to everyone, though for entirely different reasons. The two found themselves aligned. The opposition is more than twenty major tech companies: Nvidia, AMD, cloud platform operators. Their logic is straightforward. The more open-source AI gets deployed, the more enterprises buy GPUs, lease cloud infrastructure, and run local inference. Restricting open source means shrinking your own market. Four groups, four directions: the government wants to hold the gate, China wants to use price to dismantle it, Nvidia will collect on either side of it, and Anthropic believes the gate should only be open to a few. No one here is the good guy. Once you can see what each party is protecting, the headlines about AI regulation start to make considerably more sense.