Sam Altman has said the U.S. is getting it wrong on China’s AI game. The OpenAI CEO told a few reporters in San Francisco that Donald Trump’s latest ban on advanced chips won’t stop China from pushing forward in artificial intelligence.
According to CNBC, Sam met journalists over Mediterranean food in the Presidio and laid out how bad the situation really is. “I’m worried about China,” he said.
He warned that this isn’t some race where one side pulls ahead and wins. “There’s inference capacity, where China probably can build faster,” Sam said. “There’s research, there’s product; a lot of layers to the whole thing.”
He explained that the U.S. is focused on just one piece, blocking chip exports, while China is working on the entire stack. And despite growing controls, Sam doesn’t think any of it is actually helping. “My instinct is that doesn’t work,” he said when asked if fewer GPUs reaching China would be a win.
Trump, now back in the White House, signed off on a full stop of high-end chip exports to China in April 2025. That went even further than the earlier rules from President Joe Biden, which had already added restrictions to slow China’s access to advanced AI hardware.
Trump’s new policy stopped even modified chips, the ones designed to pass Biden’s rules, from going out. But just last week, Washington changed the rules again.
Under a new deal, Nvidia and AMD can now sell certain “China-safe” chips again. But they have to give 15% of their China revenue from those sales to the U.S. government. Sam didn’t talk about that deal directly, but he made it clear that trying to manage AI progress through policy alone isn’t realistic.
“You can export-control one thing, but maybe not the right thing… maybe people build fabs or find other workarounds,” he said. “I’d love an easy solution,” he added. “But my instinct is: That’s hard.”
He made it clear that this isn’t just about the chips. Chinese firms are already working with their own suppliers like Huawei. The controls from Washington haven’t stopped China. If anything, they’ve made Chinese companies move faster. While the U.S. stays dependent on Nvidia and AMD, Chinese labs are building alternatives.
Sam said that China’s progress in AI also pushed OpenAI to start opening up its own models. For years, the company refused to release full models, choosing to lock everything behind APIs. But now, with China releasing more open tools like DeepSeek, OpenAI is changing direction.
“It was clear that if we didn’t do it, the world was gonna head to be mostly built on Chinese open source models,” Sam said. “That was a factor in our decision, for sure. Wasn’t the only one, but that loomed large.”
OpenAI just released two new models: gpt-oss-120b and gpt-oss-20b. These are its first open-weight models since GPT-2 dropped in 2019. The new ones aren’t fully open source, the training data and source code are still locked, but the weights are now public.
That means developers can download and run them, even offline. The goal, Sam said, was to support people building locally-run coding agents.
Sam admitted these models won’t blow anyone away. Some developers say they’re missing key features. He didn’t argue. The team built them for one thing only, he said, and if needs change, they’ll adjust. “If the kind of demand shifts in the world,” he said, “you can push it to something else.”
Right now, OpenAI is the only major U.S. company going this route. Meta had gone open with its Llama models before, but CEO Mark Zuckerberg recently said they might stop. That puts OpenAI out front, at least for now, as Chinese labs keep releasing flexible tools that anyone can use.
Sam once said that locking up models had put OpenAI “on the wrong side of history.” This new decision looks like an effort to fix that. But it’s also a way to keep developers inside OpenAI instead of losing them to Chinese labs offering more freedom.
Get seen where it counts. Advertise in Cryptopolitan Research and reach crypto’s sharpest investors and builders.