Are the US and China entering an AI Cold War?
The only consensus in most discussions about the artificial intelligence boom seems to be that the US and China are on the brink of a cold war.
Both superpowers have not been shy about jockeying for AI dominance in recent years.
US and Chinese officials have spent more and more time focusing on semiconductor, critical mineral and export policies to try to ensure an AI victory in the years ahead.
Most recently, China seems to have caught up in the crucial area of AI model performance, and US technology analysts, politicians and technology executives seem unsure what to do about it.
In early 2025, the Chinese AI platform DeepSeek rivalled the offerings from OpenAI, Anthropic and Google.
Last week, Beijing-based start-up Moonshot AI’s open-source model Kimi K3 impressed and also scared US technology officials.
Some American officials have accused China-based firms of a practice often referred to as distillation, which is a polite way of saying that Chinese companies have potentially stolen proprietary AI technology from US firms.
On Tuesday, during an appearance on CNBC, US Treasury Secretary Scott Bessent said that the White House was considering imposing sanctions on China and on several AI models from the country. This would limit usage of those models in the US.
“We are finding watermarks of our US large language models on many of the Chinese models, and that’s unacceptable,” Mr Bessent said.
Chinese companies have denied many of the accusations, while Chinese executives and officials have made allegations about US firms.
In 2025, several Chinese media outlets reported that government officials were concerned about “backdoor features” in US semiconductors that would allow Chinese AI models to be compromised.
US AI and semiconductor companies at the centre of those reports, such as Nvidia, have repeatedly denied the claims.
Regardless, paranoia between the two countries has continued and in recent months escalated as US technology stocks experienced volatility due to Chinese AI product announcements.
That paranoia has prompted finance media personalities such as CNBC’s Jim Cramer to campaign against Chinese AI technology.
“We must not let our companies use these Chinese models to save a few bucks,” he posted on X on Tuesday, referring to open-source AI models that appeal to many because they are cheaper to operate.
“This is vital national security,” he continued. “I respect the Chinese people greatly but these companies are run by the People’s Liberation Army for heaven’s sakes.”
Whatever transpires from the US and China in terms of policies to gain the upper hand, the situation could potentially impact businesses and organisations that have spent millions of dollars on AI tools.
Florian Douetteau, chief executive and co-founder of Dataiku, a data science and AI platform company, sees the deepening US and China fracture as a wake-up call for companies that do not own the AI upon which they depend.
Mr Douetteau said he tells his clients to make sure they keep track of which AI tools they use, where they originate from and how exposed those tools are to potential geopolitical conflict.
“You have to test out hypothetical scenarios and how AI use might be impacted,” he said, comparing it to the in-depth protection in place at large banks in case computer infrastructure is hit by unforeseen events such as sanctions or tariffs.
“They often have laptops hidden somewhere in their big vaults so they can manage payroll and other items in case everything falls apart,” he said.
Mr Douetteau added that, given the climate between the US and China, it would be advantageous for businesses to ensure they “rely on local compute and consider the possibility of open-source AI models”.
Proponents say open-source models, where the code and technical specifications are freely available, democratise AI.
US companies such as OpenAI and Anthropic, however, do not follow the open-source approach, favouring a more tightly integrated and proprietary architecture.
Both companies have come under intense scrutiny from President Donald Trump’s White House, which has recently sought to exert control over products they release due to concerns about cybersecurity.
That recent pressure further complicates how the US should respond to Chinese open-source models that have the potential to dominate American markets.
Mr Douetteau said that uncertainty means businesses and organisations need to go above and beyond with preparation.
“They shouldn’t be in a situation where they’re caught in the geopolitical AI crossfire,” he said.
“Business leaders really need to look at it through the lens of strategy, autonomy and also be aware of their current operational footprint.”