OpenAI calls for global ban on DeepSeek and other Chinese AI models

OpenAI has taken an aggressive stance against DeepSeek and other Chinese AI developers, warning the U.S. government about the security risks posed by “state-subsidized, state-controlled” artificial intelligence models from the People’s Republic of China (PRC). In its policy submission for the Trump administration’s AI Action Plan, OpenAI urged the U.S. to enforce export controls that would effectively blacklist PRC-backed AI models from use in key global markets.

Citing DeepSeek’s recently launched R1 model, OpenAI drew parallels between the AI startup and Chinese telecom giant Huawei. The company warns that DeepSeek could be compelled by Chinese authorities to manipulate AI outputs, engage in surveillance, or compromise user data. This echoes long-standing concerns about Huawei’s alleged ties to Beijing, which led to its ban in multiple countries over cybersecurity fears.

OpenAI’s policy submission claims that DeepSeek’s AI is not just a technological competitor but a geopolitical threat. The document asserts that DeepSeek’s models have been trained using methods that allow them to generate content related to illicit activities, such as identity fraud and intellectual property theft — practices that align with the CCP’s disregard for American IP protections.

To counter China’s AI ambitions, OpenAI is urging the U.S. government to adopt an export control framework that would divide global AI markets into three tiers:

1. Tier I: Countries committed to democratic AI principles would receive unrestricted access to U.S. AI models and infrastructure.

2. Tier II: Nations with a history of poor IP protections or indirect PRC ties would face limited access.

3. Tier III: China and its close allies would be fully restricted from accessing American AI technologies.

OpenAI’s proposal also includes a provision for banning the use of Chinese-developed AI models in critical sectors, particularly where user privacy and security are at risk. This would include prohibiting the use of PRC-produced AI tools in government operations, data centers, and sensitive infrastructure projects.

While OpenAI acknowledges that the U.S. currently leads in AI, it warns that China is rapidly closing the gap. The company points to China’s centralized resource allocation, state-subsidized AI development, and geopolitical influence through the Belt and Road Initiative as advantages that could tilt the scales in Beijing’s favor.

The report highlights concerns over regulatory “arbitrage,” arguing that individual U.S. states adopting strict AI regulations could inadvertently disadvantage American firms while China continues to operate with looser rules. OpenAI suggests that federal preemption of state AI laws could help counterbalance this risk.

OpenAI’s submission to the Trump administration marks a significant escalation in the AI arms race between the U.S. and China. While OpenAI has previously accused DeepSeek of distilling knowledge from its models, the latest claims introduce a national security dimension, framing PRC AI as a direct threat to democratic governance and economic stability.

Whether the U.S. government adopts OpenAI’s recommendations remains to be seen. However, if implemented, the proposed AI export controls could reshape the global AI landscape, isolating China’s AI ecosystem and reinforcing America’s technological dominance.