OpenAI has escalated tensions in the global artificial intelligence landscape by advocating for a ban on AI models developed by Chinese lab DeepSeek. The American AI leader has labeled these models as “state-subsidized” and “state-controlled,” citing significant security concerns for critical infrastructure and high-risk sectors.
According to OpenAI, China’s data laws could potentially force DeepSeek to share sensitive information with the Chinese government, creating what they describe as substantial threats to national security and privacy.
AI Export Controls and the Rise of DeepSeek
DeepSeek, established in 2023 by founder Liang Wenfeng, has quickly made a name for itself with innovative AI offerings, particularly its R1 reasoning model.
The company has attracted attention for providing competitive pricing compared to OpenAI’s ChatGPT products. Despite being relatively new to the market, DeepSeek has already created significant disruption in the AI industry and gained notice in U.S. stock markets.
Adding to concerns about government connections, Wenfeng recently met with Chinese President Xi Jinping, intensifying speculation about potential ties between DeepSeek and Chinese authorities.
OpenAI has drawn parallels between DeepSeek’s AI models and previous security concerns surrounding Huawei’s telecommunications equipment. The company is urging the U.S. government to consider implementing a ban on DeepSeek’s models, particularly in “Tier 1” countries identified under the Biden administration’s export policies, including the UK, Canada, and Germany.
Simultaneously, OpenAI has proposed easing certain export controls on advanced computing chips and AI model weights.
Current Biden administration restrictions limit free export of AI chips to just 18 countries, primarily in Western Europe along with Canada, Australia, and Japan. OpenAI suggests expanding free export status to all nations that commit to deploying AI in ways that promote democratic principles and citizen freedoms.
The company has also been vocal in support of copyright laws that are AI-friendly, with emphasis on “fair use” exceptions to maintain American ingenuity. OpenAI argues that limiting AI training data to public domain content would significantly weaken model performance and competitiveness.
This has been met with criticism from copyright holders who have sued OpenAI for using content without a license.
Google and OpenAI Shared Policies, Diverging Interests in the AI Race
Google has aligned with OpenAI on several policy positions, such as open copyright policies. Google contends that there is a need to preserve “text-and-data mining exceptions” so that AI companies can train on copyrighted content effectively.
Google has been more assertive in its export control notices, however, and has contended that the tiered system adopted by the Biden administration would undermine U.S. economic competitiveness by unfairly penalizing cloud service providers.
Aside from that, Google has urged the U.S. government to invest more in long-term AI research instead of cutting federal spending and suggests making public data sets available to spur commercial AI development.
Both technology behemoths have pushed back against AI liability legislation on the grounds that the persons who implement AI systems, as opposed to their creators, are best positioned to evaluate the risks involved in them. This is within larger industry opposition to unintended side effects of excessive responsibility for innovation.
This request for limits on DeepSeek constitutes a sharp escalation of the worldwide AI race, with US businesses calling for tighter controls on Chinese AI players. As the realm of artificial intelligence keeps on shifting, these actions serve to identify the intricate entanglements among technological progress, national security interests, and geopolitics in charting the path of AI hegemony.
