The US could be overtaken by China in the global AI arms race. Could this be the wake-up call the US needs to develop more accessible, lower-cost AI models? Or are cheap AI entrants ‘too good to be true’?

This turning point, labelled a ‘warning’ by some in the media, stems from findings reported by Microsoft, which indicate that Chinese AI companies such as DeepSeek are overtaking US counterparts in markets outside the Western world. 

This data reflects a widening global shift in AI adoption and competition, particularly in price-sensitive regions. For instance, DeepSeek’s market share hit 89% in China, 56% in Belarus, and 49% in Cuba by the end of 2025.

Will China’s approach prove effective?

There are several reasons why it could. Firstly, while the US tends to innovate faster in the early stages, China scales relentlessly.

In China, AI is treated as national infrastructure. It is deeply integrated across centralised super-apps, payments, surveillance, logistics, and manufacturing, resulting in extensive real-world deployment at population scale. 

Additionally, while the US attracts elite AI talent, it can be argued that China produces vast volumes of highly competent engineers concentrated in a smaller set of companies. Its extensive STEM pipeline ensures that AI graduates move directly into implementation roles, not solely into research labs.

From the outside, China is also more willing to tolerate higher error rates and operate within what some would call ethical grey zones, whereas the US functions under stricter accountability frameworks and regulatory constraints, such as ISO 42001. Whilst necessary, these inevitably slow deployment. Such standards should ultimately be applied globally, but they do drag on competitive speed.

Whilst US firms, such as OpenAI and Google, focus on high-end, subscription-based offerings, Chinese competitors have prioritised low-cost or free models that are readily adopted by governments, startups, and local developers in emerging markets. The perceived competitive advantage is further amplified by state support, allowing Chinese LLM’s to scale rapidly.

Why DeepSeek’s rise could catch US firms off guard

The release of DeepSeek’s R1 model marked a clear turning point in accelerating global adoption of the Chinese LLM. Part of this is the strength of the model, but its free-to-use, open model has dramatically lowered barriers to adoption, accelerating its uptake. 

Its rapid rise positions it as a potential serious alternative to well-funded Western LLMs such as ChatGPT and Gemini.

Many have argued that DeepSeek’s R1 stands out for its advanced reasoning capabilities through self-reinforced learning, enabling complex problem-solving in areas such as mathematics and coding. It can provide transparent, step-by-step reasoning through chain-of-thought outputs, all at a significantly lower cost to run than Western counterparts.

Many US providers have assumed that users would consistently favour higher-quality, paid offerings. In practice, paid access is not viable in many markets. One potential response would be for US firms to explore broader free or low-cost tiers to expand adoption.

The West’s mature, measured approach to AI

That said, Western platforms retain advantages. ChatGPT, particularly at subscription levels, continues to deliver stronger multimodal functionality, a more mature user experience, and fewer concerns around censorship and data governance tied to Chinese data infrastructure. AI accountability remains one of the most critical differentiators.

While DeepSeek can excel in technical tasks such as coding and reasoning, ChatGPT, particularly within its subscription tiers, continues to offer;

• Superior multimodal capabilities
• A more refined user experience
• Fewer concerns around censorship and data privacy related to Chinese servers

DeepSeek may be challenging the US’s big tech dominance and demonstrates that advanced AI development can be achievable with fewer resources, a particularly significant outcome for China amid US chip restrictions. The question, then, is whether current restrictions are containing competitors, or are they quietly accelerating their independence?

The hidden risks behind ‘free’ AI

Enterprises, however, may underestimate the security, accountability and governance risks associated with adopting new Chinese LLMs, often increased by unsanctioned employee usage, resulting in shadow AI.

DeepSeek operates as a shared cloud service with data stored in China, introducing potential risks for organisations whose data may be shared unknowingly by employees. While these risks can be mitigated through local hosting, few organisations are willing to accept the associated complexity, cost, and ongoing maintenance burden.

Recommended reading

• China’s Qwen Hits 700m Downloads to Dominate Open-Source AI
• Is Europe Already Too Far Behind in the AI Race?
• ChatGPT Misuse Drives Majority of Enterprise AI Data Exposure

Concerns around data privacy, regulatory compliance, transparency and accountability are central for businesses and governments alike, and DeepSeek’s rapid adoption represents a new level of exposure for both.

Leadership teams must treat AI adoption, including the use of LLMs, generative AI, agents, etc., as a strategic risk decision, not merely a cost-saving opportunity. It is important that organisations apply rigorous scrutiny to all AI services, not just Chinese models. 

Privacy policies, data residency, security controls, and transparency must be core evaluation criteria and cannot be bolted on after deployment.

US leadership depends on AI governance, the real ‘AI arms race’

Despite the media’s warning that China could be overtaking the US, the AI race is not solely about processing power, model performance or affordability. Trust, governance, interoperability and accountability are decisive differentiators.

Nations that embed responsible AI principles early will shape global standards. Robust AI governance underpins trust with international partners and enterprise customers alike. ISO 42001, especially, is an emerging international standard for AI management systems, and these frameworks provide a structured way to embed accountability. Maintaining accountability by design is non-negotiable.

The US government’s approach to AI standards will ultimately determine its long-term competitive position, particularly its ability to sell into Europe and other regulated markets.

So, should the US be worried? Well, it depends on where you stand. If US companies aspire to remain dominant in the global AI race, producing cheaper, lower-quality models for emerging regions is a strategic avenue worth serious consideration, but not by compromising trust, security and accountability.