Chinese AI models surge globally, raise concerns in US report
Chinese AI models are seeing rapid global adoption, raising concerns in the US over data security, supply chains and economic impact. A report highlights risks including intelligence access, weaker safeguards and growing competition for American firms in global markets
New Delhi: Chinese artificial intelligence (AI) models are rapidly gaining global traction, raising concerns among US policymakers over potential risks to national security, supply chains, and economic competitiveness, according to a report.
As per the War on the Rocks analysis, Chinese AI systems, which accounted for just 1 per cent of global workloads in late 2024, surged to nearly 30 per cent by the end of 2025.
Models developed by companies such as Alibaba, along with emerging players like DeepSeek, Moonshot AI, and MiniMax, are increasingly being used in applications ranging from academic research to enterprise solutions.
The report highlighted that while these models are often open-source and freely available, they are developed under China’s legal framework, which requires firms to cooperate with national intelligence efforts, raising concerns over potential data access by authorities.
Experts cited in the report warned that the widespread use of such models could expose sensitive information, as users increasingly rely on AI systems for proprietary code, business strategies and confidential communications.
The analysis identified four major areas of concern for the US, such as risks to AI supply chains, potential intelligence collection, enhanced capabilities for malicious actors, and long-term economic displacement.
One key issue flagged was the difficulty in auditing AI models for hidden vulnerabilities or ‘backdoors’, especially in open-source ecosystems.
Moreover, security researchers have found that compromised datasets can embed malicious instructions that are extremely hard to detect during standard checks.
The report also pointed out the risks of data exposure when AI systems process user inputs through servers located in China, particularly in cases where developers integrate such models via application programming interfaces (APIs).
In addition, Chinese AI models were found to have weaker safety guardrails compared to their Western counterparts, making them more susceptible to misuse, including generating harmful code or aiding cyberattacks.
On the economic front, the growing popularity of low-cost Chinese models could challenge the dominance of US firms, especially in price-sensitive markets and developing regions, where they are increasingly becoming the default choice.
The report indicated that instead of outright bans, policymakers should focus on transparency, minimum safety standards, and supply chain accountability to mitigate risks while maintaining competitiveness.
