U.S. Lawmakers Navigate Challenges with DeepSeek AI Startup

The DeepSeek AI application is captured in this illustration taken in Warsaw, Poland, on January 27, 2025.

U.S. lawmakers are closely watching the ascent of DeepSeek, a Chinese artificial intelligence startup that could potentially influence the American AI landscape. Experts are concerned about the startup's rapid progress and its implications for existing U.S. trade restrictions.

DeepSeek's emergence raises questions about how effective the U.S. government's efforts are in limiting China’s ability to access advanced technology for national security reasons. This situation has federal lawmakers contemplating new strategies to address what some are calling a serious threat posed by DeepSeek.

DeepSeek has capitalized on open-source technology, which makes it challenging for the U.S. government to impose export restrictions effectively. According to Lawrence Ward, a partner at the law firm Dorsey & Whitney, while the company might face civil or criminal penalties if it improperly uses specific Nvidia chips, enforcing these penalties could be extremely difficult.

On a recent Monday, lawmakers urged action to hinder DeepSeek’s progress, emphasizing the need for stronger restrictions given the rapid advancements the company is making. The model is particularly concerning due to its unique capabilities, which some believe surpass those of well-established American AI firms.

Industry experts like Paul Triolo highlight that creating regulations to limit an open-source model’s use is complicated, particularly because such technologies are already popular among various organizations and users. One potential route could involve the Commerce Department imposing rules that would require major technology firms, such as Apple and Google, to remove DeepSeek's application from their platforms, though eliminating the app from repositories like Github would remain a significant hurdle.

DeepSeek recently introduced a free, open-source large language model that was developed in only two months and reportedly at a much lower cost than similar products from major competitors. In a remarkable achievement, the company's latest reasoning model has been said to outperform even OpenAI's recent offerings in several independent assessments.

Despite imposed controls by the U.S. over China's access to premium Nvidia chips in 2022, DeepSeek, based in Hangzhou, claims to have developed its model using less advanced Nvidia H800 chips. These chips were available before the U.S. extended its ban in October 2023.

As the situation unfolds, experts agree that the U.S. is facing unprecedented challenges. Paul Triolo states, "We are clearly in uncharted waters at many levels." Daniel Newman, CEO of Futurum Group, also points out that if DeepSeek has acquired tools against regulations, it would prompt further investigations into the company's operations.

U.S. President Trump noted that the success of the DeepSeek AI app should serve as a significant alert for American technology firms. He praised the company's ability to deliver effective solutions at a lower cost, suggesting that there are lessons to be learned from its approach to AI development.

DeepSeek's cost-effective solutions have raised concerns among investors, causing a ripple effect in U.S. tech stocks, as people began to question the substantial investments larger firms are making in AI and related infrastructures. Companies in the U.S. could also utilize DeepSeek's public codes to enhance their own AI models, potentially lowering their operational costs.

Representative John Moolenaar, who chairs the House Select Committee on China, emphasized the urgency for the U.S. government to implement more stringent export controls on critical technologies related to DeepSeek's capabilities. He remarked on the uncharted nature of the situation and questioned whether anyone in Washington has clear directives moving forward.

In response to a recent large-scale cyberattack that disrupted its services, DeepSeek has restricted new sign-ups to users with phone numbers registered in mainland China. The timing of this move coincides with the launch of their R1 reasoning model, which attracted significant attention, including participation by DeepSeek’s founder in a high-level government symposium led by the Chinese Premier.

DeepSeek, AI, Trade