Several major US artificial intelligence companies have expressed fears about erosion of the US edge in AI development.
In a recent submission to the US government, companies warned that Chinese models such as the Deepseek R1 are becoming more refined and more competitive. Submissions submitted in March 2025 in response to requests for input into the AI Action Plan highlight the growing challenges from China in technical capabilities and pricing.
The presence of Chinese AI
China’s national support AI model Deepseek R1 has piqued the interest of US developers. According to Openai, Deepseek shows the narrowing of the technological gap between the US and China. The company describes DeepSeek as “states subcissidized, state-controlled and freely available,” raising concerns about the model’s capabilities to impact global AI development.
Openai compares Deepseek to Chinese telecommunications company Huawei, warning that Chinese regulations can force the government to compromise on Deepseek-sensitive US systems or infrastructure. Concerns have also been raised about Data Privacy, with Openai noting that China’s rules could force Deepseek to disclose user data to the government, increasing China’s ability to develop more advanced AI systems.
The competition with China also includes Baidu-released Ernie X1 and Ernie 4.5, designed to compete with Western systems.
According to Baidu, the Ernie X1 “provides performance at half the price in parallel with the Deepseek R1.” Meanwhile, the Ernie 4.5 priced just 1% of Openai’s GPT-4.5, surpassing that on multiple benchmarks.
Deepseek’s aggressive pricing strategy has also raised concerns for US companies. According to a Bernstein research, Deepseek’s V3 and R1 models are “20-40 times cheaper” than Openai’s equivalent. Pricing pressures could force US developers to adjust their business models to stay competitive.
Baidu’s strategy of open-sourcing models has also gained traction. “One of the things I learned from DeepSeek is that open-sourcing the best models can really help you adopt,” Baidu CEO Robin Li said in February. Baidu plans to open source the Arnie 4.5 series from June 30th.
Aside from the cost, early user feedback on Baidu’s model is positive. “(I’ve been playing with it for hours,” said Alvin Foo, a venture partner at Zero2Launch, in a social media post, suggesting that China’s AI models are becoming more affordable and effective.
Security and economic risks of US AI
The submission also highlights what US companies perceive as risky to their safety and economy.
Openai warned that Chinese regulations could allow Deepseek to manipulate models to compromise infrastructure and sensitive applications, creating vulnerabilities in critical systems.
Human concerns are concentrated on biosecurity. It revealed that the unique Claude 3.7 Sonnet model demonstrated its capabilities in the development of biological weapons and emphasized the dual use of AI systems.
Humanity has also raised issues regarding US export control of AI chips. Nvidia’s H20 chips meet US export restrictions, but still work well with text generation. This is an important feature of reinforcement learning. Humanity has called on the government to step up controls to prevent China from using chips to gain a technical edge.
Google took a more cautious approach, acknowledging security risks while warning against overregulation. The company argues that strict AI export rules could undermine the US’s competitiveness by limiting business opportunities for domestic cloud providers. Google recommended targeted export restrictions to protect national security, but there was no disruption in its operations.
Maintain AI competitiveness
All three US companies emphasize the need for government surveillance and improved infrastructure investments to maintain US leadership.
By 2027, humanity warned that training a single advanced AI model might require up to 5 gigawatts of power, sufficient to power small cities. The company proposed a national goal of creating 50 additional gigawatts of AI-only power capacity by 2027 and streamlining regulations on power transmission infrastructure.
Openai positioned the competition between the US and China AI as a contest between the AI models of democracy and authoritarianism. The company argued that by promoting a free market approach, it would promote better outcomes and maintain America’s technological advantage.
Google focused on requests for practical measures, including increasing federal funding for AI research, improving access to government contracts, and streamlined export controls. The company also recommended more flexible procurement rules to accelerate adoption of AI by federal agencies.
US AI Regulatory Strategy
US companies have sought a unified federal approach to AI regulation.
Openai proposed a regulatory framework controlled by the Commerce Department, warning that fragmented state-level regulations could drive AI development overseas. The company supported a layered export management framework, providing broad access to democratic US development AI while restricting it in authoritarian countries.
Humanity has called for stricter export controls on AI hardware and training data, warning that mild improvements in model performance could potentially provide China with strategic advantages.
Google focuses on copyright and intellectual property rights, emphasizing the importance of interpreting “fair use” for AI development. The company warned that excessively restricted copyright rules could be at a disadvantage for US AI companies compared to their Chinese competitors.
All three companies highlighted the need for faster government adoption of AI. Openai recommended removing existing testing and procurement barriers, but humanity supported a streamlined procurement process. Google highlighted the need for improved interoperability in government cloud infrastructure.
See: Best AI Prompt Generator: Create the perfect AI Prompt
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London. The comprehensive event will be held in collaboration with other major events, including the Intelligent Automation Conference, Blockx, Digital Transformation Week, and Cyber Security & Cloud Expo.
Check out other upcoming Enterprise Technology events and webinars with TechForge here.
