Late last month, the US Treasury Department finalized new restrictions limiting what kinds of Chinese tech startups US venture capital firms can invest in for national security reasons. When they go into effect in January, the long-awaited measures will stop American VCs and other investors from pouring money into cutting-edge Chinese AI models. After president-elect Trump takes office a few weeks later, his administration may expand the rules and make them even tougher.
While the US still leads the world in advanced AI development, the American government has grown increasingly concerned about China catching up soon. The new outbound investment restrictions are designed to work alongside other measures, such as export controls on advanced computer chips and the Committee on Foreign Investment in the United States (CFIUS), to collectively hamper—or at least slow down—the progress of Chinese AI companies.
The new restrictions outline two broad no-go zones for US AI investors. The first is companies developing tech designed exclusively for sensitive uses like China’s military and intelligence services, which investors will be prohibited from putting money into regardless of how advanced their AI capabilities may be.
For Chinese startups working on things like consumer AI, American VC firms will need to use a specific threshold to determine whether they can make investments in them. A startup’s AI model can be no larger than 1025 flops, or floating point operations per second, a numerical measure that indicates the size and capabilities of an AI model. (The European Union used the same threshold to determine which AI systems should be subject to greater regulatory scrutiny.) A lower threshold, 1024 flops, will be applied to AI systems that are primarily trained using biological sequence data due to fears that such tools could be used for particularly sensitive purposes, like creating bioweapons.
“A 1025 [flops] model is the frontier. The largest models that exist today all exist in this band between 1025 to 1026 [flops],” says Jaime Sevilla, the director of Epoch AI, a research organization that tracks the trajectory of the AI industry. That includes all the big names like OpenAI’s GPT, Google’s Gemini, Facebook’s Llama, and Anthropic’s Claude. “A 1024 flops model is what the frontier was like in 2022. It is much more common for most to be training above the 1024 flop [threshold nowadays],” Sevilla explains.
Epoch AI is keeping track of the flop counts that AI companies have disclosed publicly or can be estimated from their technical specifications, though Sevilla says that firms are becoming more discreet about revealing the exact measurements of their models. For now, there are only two Chinese companies, TikTok owner ByteDance and Zhipu AI, that have publicly announced models exceeding 1025 flops. When it comes to Chinese models trained primarily on biological sequence data, none have surpassed 1024 flops.
Emily Kilcrease, a senior fellow at the think tank Center for a New American Security, says that the Treasury Department’s 1025 flops threshold will be easier for the US government to enforce than the blanket ban on all investments in AI systems with military applications. “Given AI is fundamentally dual-use, I always view [end use restrictions] as fundamentally unworkable,” she says. “They’re kind of a complement to what I actually think are the more workable restrictions based on the computing power of the AI model.”
By setting the 1025 flops standard, the Treasury Department is targeting a relatively narrow swath of China’s AI sector, which will immediately be cut off from any potential US investments in January. Given their small scope, the restrictions may have limited immediate financial impact, but they will likely help discourage all Chinese AI companies from seeking help, financial or otherwise, from the United States in the future. That may wind up being especially true if the incoming Trump administration decides to make them even stricter.
Several of Trump’s early appointees have a long track record of being tough on China, and some have specifically expressed interest in curbing the flow of investor money from the US to the People’s Republic. Marco Rubio, the US senator from Florida who has been nominated to become Trump’s new secretary of state, proposed a bill in September that would significantly raise the amount of taxes that investors have to pay on their investments in Chinese companies.
More Homework
In the short term it’s hard to say whether the Chinese AI industry will feel the impact of the new rules. Investment flows between the US and China have already been reduced to a trickle following years of tense relations between the two countries, particularly on issues related to advanced technology. Regulatory uncertainties, the weakened Chinese economy, and China’s domestic tech crackdown all contributed to the decline in investor activity as well.
“The high-water mark of these types of investments was really about 2016. And since then, there’s been a pretty big decline,” says Sarah Bauerle Danzman, an associate professor of international studies at Indiana University, Bloomington, who previously assisted the US government in reviewing foreign investments. But there’s a chance that American investors become interested in making deals with Chinese AI firms again, and the new restrictions are “really seen as battening up the hatches even if we don’t think that there’s a real deal deluge coming through,” Danzman says.
The more certain short-term impact is that US investors who are still interested in Chinese AI startups will have to do a whole lot more due diligence. The Treasury Department is not setting up a new government committee like CFIUS that will review every transaction investors submit, and is instead asking them to do their own homework and report whether they believe a Chinese AI company would be covered.
Under the new rules, even if a Chinese startup’s AI model is smaller than the 1025-flops size threshold, a US investor might still have the responsibility to notify the Treasury Department about their transaction and the homework they’ve done, as long as its model is at least 1023 flops (essentially encompassing all large-scale models being developed today and in the future). In effect, that means the US government is creating its own system to monitor the overall flow of money going from US investors to Chinese companies working on AI.
“In order to confirm that a transaction is out of scope, it will require significant due diligence undertaken by US investors,” says Robert A. Friedman, an international trade lawyer at law firm Holland & Knight. While the rules have been celebrated by domestic AI companies and their backers, they will become a hurdle for venture capitalists with international portfolios, he says.
Uncertain Future
The outbound investment restrictions are set to take effect on January 2, and in the meantime, the Treasury Department has signaled that some small changes are still on the way to further clarify the rules. Officials also said they’re making efforts to coordinate with US allies, like the G7 countries, to introduce similar measures that would prevent Chinese AI companies from turning to VCs in Europe, Canada, or Japan for the kinds of investments prohibited in the US.
The biggest uncertainty now, as with most parts of the US federal government, is how a second Trump presidency might change things. Danzman notes that many members of the venture capital community that supported Trump are against the kinds of regulations introduced by the Treasury Department, so they may potentially try to lobby the president to roll them back. Several major American companies, like Tesla and Blackstone—both led by outspoken pro-Trump billionaires—have significant investments in China and could see their businesses negatively impacted by tighter constraints.
Other experts WIRED spoke to expect the new Republican administration, which is slated to include a number of China hawks like Rubio, will expand the scope of the rules. “It is possible that we could see a new executive order. Or, given the unified Republican government, perhaps expansion would take place via legislative action,” says Kilcrease. That might mean more measures targeting other kinds of Chinese startups, in sectors ranging from biotechnology to batteries.
The Biden administration’s tech policy toward China has been defined by, at least in principle, the idea of a “small yard, high fence,” or in other words, designating relatively narrow areas where the US government can set very strict restrictions. The latest version of the outbound investment rules is an example of what that idea looks like in action. But under Trump, Chinese companies might end up seeing just how large the yard can actually get.