OpenAI calls DeepSeek ‘state-controlled,’ calls for bans on ‘PRC-produced’ models

OpenAI claims DeepSeek is state-controlled, calls for banning PRC-produced models.

: OpenAI describes DeepSeek as a 'state-subsidized' and 'state-controlled' AI lab from China, advocating for banning its models to mitigate security risks. Their policy proposal submitted to the Trump administration argues DeepSeek's compliance with Chinese laws poses a risk of IP theft and data demands. Unclear classification of DeepSeek's models and allegations suggest heightened tension between OpenAI and DeepSeek. Despite no clear link to the Chinese government, China's interest surfaces as DeepSeek's founder meets Xi Jinping.

OpenAI, a leading artificial intelligence research organization, recently directed its attention towards DeepSeek, a well-known AI lab based in China, denouncing it as both 'state-subsidized' and 'state-controlled.' In a move that suggests further escalating tensions between the two entities, OpenAI has submitted concerns to the Trump administration's 'AI Action Plan,' proposing the U.S. contemplate banning models developed by DeepSeek and similar entities supported by the People’s Republic of China (PRC). OpenAI argues these models, including DeepSeek's R1 'reasoning' model, are fundamentally insecure given DeepSeek's legal obligations under Chinese law to share user data.

OpenAI's proposal speaks to the inherent risks in using 'PRC-produced' models, identifying potential threats in the form of privacy violations and high risks of intellectual property theft. The call for a ban extends to all nations considered 'Tier 1' according to the export rules established by the Biden administration. Although these models are renowned for their efficiency, OpenAI holds that their adoption could result in compromising sensitive data and technologies.

Despite OpenAI's claims, the nuances regarding which model categories are implicated—whether APIs or open models—remain ambiguous. DeepSeek's open models, though devoid of data siphoning measures, are hosted by several major companies, including Microsoft, Perplexity, and Amazon. Previously, OpenAI had accused DeepSeek of unlawfully distilling knowledge directly from OpenAI's own models, marking a breach of terms of service. Today's accusations stand as a progressive intensification of OpenAI's wary approach to DeepSeek.

The connection between DeepSeek and the Chinese government, though implied, lacks transparent evidence. DeepSeek emerged from High-Flyer, a quantitative hedge fund, and operates independently. Nonetheless, recent developments reveal the PRC's growing interest, particularly captured by an encounter between DeepSeek founder Liang Wenfeng and Chinese leader Xi Jinping a few weeks prior. This meeting underscores the perceived significance DeepSeek holds for the Chinese government amidst a global climate of tech rivalry.

The implications of OpenAI's assertions touch on broader themes of cybersecurity and control over advancements in machine learning methodologies. While robust security measures are crucial for international collaborations in technology, the complexity of entwined geopolitical motives and corporate interests enhances the challenges. Comprehensively addressing these subjects requires transparency, due diligence, and cooperation across sectors to balance innovation with ethical practices.

Sources: Reuters, TechCrunch, The Verge, Wired, CNBC

AImachine learningcybersecurity