DeepSeek’s new AI model hasn’t just rocked America’s tech sector — the Chinese company’s success has also alarmed lawmakers, who are now searching for ways to stop the country from making further tech strides.
“We are in a global race for artificial intelligence,” Sen. Ted Cruz told reporters last week. “And there’s no doubt that China’s doing everything they can to beat the United States.”
Members of Congress from both parties agree that DeepSeek’s rapid development indicates there will be hurdles ahead in that race. Federal agencies and Hill offices also see it as a security threat, banning staff use of the product.
But beyond that, lawmakers aren’t exactly unified about how to respond. Some Republicans want to ban any transfer of artificial intelligence technology to Chinese companies, which would be possibly challenging to enforce. Others told NOTUS this week that they want even fewer regulations on the burgeoning AI industry, arguing such a move would allow American companies to innovate faster. And some said the key to success is poaching China’s brightest scientists and allowing them to work in America, rather than returning to China after studying at U.S. universities.
“One of our problems is we educate super smart people from other countries and then require them to go home,” North Dakota Sen. Kevin Cramer told NOTUS.
If DeepSeek’s claims about developing its model with fewer resources than American tech companies are true, it also may demonstrate that Chinese developers have managed to get ahead and adapt to the hardware export restrictions that America has imposed on China.
OpenAI said last week that they found evidence that DeepSeek might have exploited their developer API to train their model on top of OpenAI’s data without authorization. Researchers at OpenAI and Microsoft say they are working to confirm this claim and are yet to release their findings.
DeepSeek’s open-source model — even as it refuses to acknowledge the Tiananmen Square protests and parrots Chinese propaganda in response to many queries — rivals some of the most advanced OpenAI large language models. The company claims its AI was trained at a fraction of the cost — only $6 million, a drop in the bucket compared to the estimated $41 million to $78 million that OpenAI spent training ChatGPT-4.
“Until such time as we get our act together here in the United States,” Republican Sen. Mike Rounds summarized, “we’re gonna be in one heck of a race with China.”
Some lawmakers, however, alleged the Chinese company might be bluffing about the cost.
The official price tag is “total BS,” Ohio Sen. Bernie Moreno said. “I don’t buy the Chinese, that they were able to accomplish that.”
Cramer, meanwhile, argued it just means Americans should get more serious about the tech race.
“Maybe we’re not as advanced innovators as we thought we were,” he said. “But we can recover from that quickly.”
In a series of executive orders, President Donald Trump directed his adviser for artificial intelligence and cryptocurrency to develop a plan to secure American AI dominance. Trump also rolled back several Biden-era requirements for developers working on frontier AI models.
Sen. Josh Hawley, on the other hand, wants more limits for American tech companies. The Missouri Republican introduced legislation to ban exports of artificial intelligence technology to China, block American companies from researching AI in China (or in collaboration with Chinese firms) and ban American investors from contributing to Chinese AI efforts.
“Let’s make sure that we’re not aiding them,” he told NOTUS.
On Monday, Hawley and Sen. Elizabeth Warren released a joint statement asking the Trump administration to impose tighter chip sale restrictions. Biden-era blockages, the senators said, still allowed manufacturers like Nvidia to sell chips powerful enough to train AI to Chinese developers.
Sen. Steve Daines, though, feels strongly that “government is not going to be the solution here.”
“America needs to let engineers go faster with the right tax policy and regulatory regime,” he argued. “So it’s a wake-up call. We need to continue to pour gas into our innovation ecosystem and the private sector.”
Florida Sen. Rick Scott shared similar priorities. “America will always win,” he said. “Our innovation will always be better than everybody else’s.”
Trump announced that several firms will invest in AI infrastructure in a $500 billion venture known as Stargate, building data centers to offer researchers more capacity.
A bipartisan commission that tracks economic and security developments in China has argued for a large-scale effort. The panel recommended in its 2024 report that Congress “establish and fund a Manhattan Project-like program dedicated to racing to and acquiring an Artificial General Intelligence (AGI) capability.”
Melanie Hart — the director of the Atlantic Council’s Global China Hub, who worked on semiconductor policy while at the State Department — argued during a hearing last week, like Cramer, that America’s best bet is to siphon off China’s most talented engineers and to keep bright foreign students after they graduate.
Senate Foreign Relations Committee Chairman Jim Risch, an Idaho Republican, was skeptical of that idea. He noted that those students’ families would still be in China, where the Chinese government could threaten them and try to place pressure on Chinese workers abroad.
Hart emphasized on Thursday that the competition isn’t only about who will dominate the next tech era.
“The U.S. and China are engaged in systemic competition,” she said. “It is a competition over which model prevails: the U.S. model of political and economic development or China’s. If China prevails, the U.S. and the world will be less free, less prosperous and less safe.”
—
Samuel Larreal is a NOTUS reporter and an Allbritton Journalism Institute fellow. Haley Byrd Wilt is a reporter at NOTUS.