OpenAI Accuses DeepSeek of 'Free-Riding'
OpenAI has accused Chinese AI firm DeepSeek of malpractice, alleging the company is "free-riding" by distilling its proprietary models. The accusation highlights growing geopolitical tensions in the AI sector. The dispute raises questions about intellectual property protection and the ethics of model training in the global AI race.
- In a memo to the U.S. House Select Committee on China, OpenAI detailed alleged "new, obfuscated methods" used by DeepSeek to bypass its safeguards. The company claims it observed accounts linked to DeepSeek employees using third-party routers and developing code to programmatically access and extract outputs from U.S. AI models for distillation. - OpenAI has raised concerns that when AI capabilities are copied through distillation, the original safety features may not transfer, creating risks for misuse in sensitive areas like biology or chemistry. The company also noted that DeepSeek's chatbot appears to censor topics sensitive to the Chinese government, such as Taiwan and Tiananmen Square. - DeepSeek was founded in 2023 by Liang Wenfeng, who also co-founded the quantitative hedge fund High-Flyer. Liang has a master's degree in information and communication engineering from Zhejiang University and has stated a goal of closing the "one- or two-year gap" between China and the U.S. in AI. - Despite presenting as an independent startup, DeepSeek has significant ties to the Chinese government. Its parent company was designated a "national high-tech enterprise," granting it preferential tax treatment and subsidies, and founder Liang Wenfeng was invited to a symposium with Chinese Premier Li Qiang. - Investigations allege that DeepSeek has received support from the Chinese Communist Party (CCP) and has connections to the People's Liberation Army (PLA). One report identified 396 AI research projects funded by the PLA that involved DeepSeek-affiliated researchers. - The dispute occurs as the performance gap between U.S. and Chinese AI models narrows significantly. According to one rating system, the lead held by U.S. models shrank by 78% between January 2024 and February 2025. - This accusation is part of a larger pattern of intellectual property disputes in the AI industry, with numerous lawsuits filed against AI developers by news organizations, authors, and artists over the use of copyrighted material for training models. - The incident highlights the intense U.S.-China competition in AI, a key part of China's national strategy to become the world leader in AI by 2030. In response to China's advancements, the U.S. has implemented export controls on advanced AI chips to slow its progress.
