US ban on DeepSeek may deter private firms from using low-cost Chinese AI

The federal ban on DeepSeek

The US Department of Homeland Security (DHS) recently issued a directive requiring federal civilian agencies to strip DeepSeek products from their IT infrastructure. This order, signed in July and made public on Monday, also prohibits these agencies from purchasing any third-party services built on the Chinese AI model.

The government is citing national security concerns as the primary reason for the move. While the order specifically targets federal civilian agencies, it serves as a formal statement on the risks associated with integrating Chinese-made artificial intelligence into government systems.

In response, China has criticized the move as a discriminatory measure that undermines fair competition. Beijing accused Washington of using national security as a pretext to politicize trade and technology, warning that it would take steps to protect its own interests.

Impact on the private sector

Industry analysts suggest that the DHS order could have a chilling effect on the broader American business community. Even though the ban applies to federal agencies, private companies that serve both government and commercial clients may now view DeepSeek as a liability. Software developers and cloud providers may now reassess their tech stacks to ensure they do not jeopardize future government contracts.

The disclosure of the order transforms an internal procurement rule into a broader market signal. Experts note that this could fundamentally change the cost calculations for American businesses when they select which AI models to power their products. Companies may feel pressured to avoid Chinese models even if they are the most efficient options available.

This regulatory pressure comes as US lawmakers look to expand these restrictions further. The goal is to create a more complex compliance environment for private-sector suppliers who provide services to the government, potentially raising the cost of doing business for many American tech firms.

The appeal of Chinese models

Despite the growing regulatory risks, Chinese AI models have gained significant traction among US developers due to their performance and low operating costs. Many American firms have turned to these models specifically to manage the high expenses associated with deploying generative AI at scale.

Several prominent US tech players have already integrated Chinese technology into their workflows. For example, Airbnb CEO Brian Chesky noted that the company uses Alibaba Group Holding’s Qwen models because of their speed and affordability. Other companies, including the legal tech firm Harvey and the conversational engine Perplexity AI, have also utilized models from DeepSeek or Moonshot AI.

For smaller developers, the price difference is especially stark. One US-based developer shared with the publication Rest of World that a coding task costing roughly $10 using Anthropic’s Claude could be completed for less than 50 cents using DeepSeek. These savings make Chinese models highly attractive for startups with limited budgets.

A multipronged US strategy

The DHS ban is part of a larger, multipronged strategy by Washington to counter the popularity of Chinese open-weight AI models. This strategy includes procurement bans, defense supply chain rules, and increased congressional scrutiny to raise the political and financial risks for American adopters.

By increasing these hurdles, the US aims to limit the influence of Chinese AI in the domestic market. However, these measures also threaten to undercut the primary benefits of Chinese models, which are valued for their flexibility and ability to be deployed on private infrastructure. The government is essentially creating a trade-off between cost-efficiency and geopolitical compliance.

The situation is complicated by the fact that the national security concerns cited by the DHS could theoretically be applied to other Chinese systems. This creates uncertainty for businesses about which specific models will remain viable in the long term as the regulatory landscape continues to evolve.

China's response and shifts

China is not remaining passive in the face of these restrictions and is moving to tighten its own oversight of AI technologies flowing overseas. Last week, Beijing proposed new export controls covering advanced AI models amid its own rising security concerns.

Some domestic industry figures in China are calling for a shift toward closed-source models. Wang Yuquan, a founding partner at Haiyin Capital, argued that open-weight releases expose Chinese developers to foreign regulatory risks while allowing overseas platforms to profit from their technology.

Transitioning to proprietary models is seen as a way for Chinese developers to retain more control over deployment. By moving away from open-weight releases, these companies hope to reduce their exposure to foreign bans, build more sustainable revenue streams, and protect their technological interests from international interference.

The path forward

The tension between cost-efficiency and national security is likely to define the next phase of AI adoption in the United States. Companies will have to decide whether the immediate savings of Chinese models outweigh the long-term risks of being locked out of government-related business.

As the US continues to refine its restrictions, the software industry will need to navigate a complex web of compliance requirements. This may lead to a bifurcated market where some companies prioritize low-cost Chinese models for private use, while others stick to Western models to ensure government compatibility.

The ultimate impact of the DeepSeek ban will depend on how quickly these restrictions expand to the private sector. If the government continues to target Chinese AI, the cost of deploying generative AI in the US may rise as companies are forced to abandon the most affordable options on the market.

What to watch

Will the US government expand the DHS ban to include private-sector suppliers who serve federal clients?

How will the cost of AI deployment for small US startups change as they are pushed away from low-cost Chinese models?