DeepSeek’s breakthroughs are too big for the US to ban

DeepSeek’s moment of global glory is a double-edged sword, and it couldn’t have come at a more geopolitically fraught time.

The upstart garnered international attention with its latest models, shocking Silicon Valley and investors with their capabilities and efficiency. But for a Chinese AI company with overseas ambitions, doing the technically impossible will amount to the easy part. If it wants to remain in the lucrative US market, the real challenges start now.

Given the fate of TikTok in the US and the recent blacklisting of Tencent Holdings Ltd., I suspect it’s only a matter of time before Hangzhou-based DeepSeek will face a sweeping US crackdown. It doesn’t help that it stores data on Chinese servers. Or that animosity between Beijing and Washington runs high as President Donald Trump has unleashed fresh tariffs as part of the opening salvo of his anticipated trade war.

Regulators from Italy to Taiwan to Texas have already started restricting the use of DeepSeek. Global scrutiny will only multiply in the coming weeks and months. The national security arguments driving America’s looming TikTok ban, which is currently experiencing a temporary reprieve thanks to Trump, are only heightened when it comes to the generative AI application. And its growing number of international users — it shot to the top of US app stores over the past week — will raise global lawmakers’ concerns that it could be used by Beijing to gather large swaths of user data or sway their opinions with its outputs.

Many Silicon Valley leaders have said they’re impressed by DeepSeek and welcome competition. But they’ve also used it to advocate for more support for US AI infrastructure and policies to further hold back China. (US officials are also probing whether DeepSeek obtained advanced Nvidia Corp. chips by circumventing restrictions).

Covering technology from Asia, I’m tempted to imagine an alternate reality where the Chinese and US consumer tech ecosystems could coexist. There is no doubt that Americans would be better off with the lower-cost and more efficient AI offerings that DeepSeek has developed. Embracing competition from China, instead of barring it, would also help address the bipartisan concerns surrounding US tech giants’ market dominance.

Yet such a techno-utopia that is mutually beneficial to both sides and allows Chinese and American tech companies to compete and complement each other, benefitting consumers, businesses and innovation, could not exist in the current geopolitical climate. As the TikTok saga has played out, US lawmakers have made clear that any Chinese tech products are considered a potential national security threat. The designation of gaming publisher Tencent as an alleged Chinese military company hammered that viewpoint home. And, unfortunately for DeepSeek, the stakes are arguably higher when it comes to AI, which lawmakers worry could have military applications.

But the developers of DeepSeek made a cheeky, consequential choice in developing their AI models open source (or in this case, so-called open weight, for those who want to be technical). They also published a research paper on how they built their model, although withholding certain details about the training data. They made their product free for global users and developers to tinker with. Perhaps most importantly, the open-source route provides a tremendous amount of transparency, potentially assuaging fears about Beijing’s control. As my colleague Parmy Olson has written, this stands in stark contrast with the route OpenAI has taken in keeping its proprietary AI models shrouded in a black box.

In other words, Washington can ban DeepSeek’s app, but the startup has published most of its breakthroughs online. It’s also allowing programmers around the world to modify it and build out new, tailored products using its technology for free. Such a service will be difficult to stamp out at this point. It also could be viewed as an ironic move from a country that ostensibly celebrates free markets and innovation. As prominent Silicon Valley investor Marc Andreessen put it: “Closed source, opaque, censorious, politically manipulative vs open source and free is not the winning position the US needs.”

There’s a valid point to be made that banning Chinese software and technology is fair game for the US, given that China prohibits most US tech influence behind the Great Firewall, from Meta Platforms Inc.’s social media sites to Alphabet Inc.’s Google search engine. Instead of following the playbook of an authoritarian regime, and clamping down on a democratizing piece of technology, the US should be strategic with its next move. Tech leaders and policymakers should examine how DeepSeek’s open-source breakthroughs could be used to drive American AI innovations.

Not allowing DeepSeek’s products in the US may be inevitable, but it will be nearly impossible for Washington to eliminate their influence. And blocking the app won’t stop the rest of the world from accessing its underlying technology. Policymakers and tech leaders should react to this moment by fully embracing the open-source movement.