OpenAI Accuses China’s DeepSeek of AI Model Distillation in Memo to Congress
OpenAI has alerted U.S. lawmakers that Chinese AI startup DeepSeek is systematically bypassing access safeguards to replicate its proprietary models. In a memo to the House Select Committee on China, OpenAI alleged DeepSeek employees used obfuscated routing and automated extraction techniques to distill outputs from ChatGPT and other U.S. systems.
The allegations come amid heightened U.S.-China tensions in frontier AI development. DeepSeek's rapid ascent since its 2025 debut has drawn scrutiny from Washington, with the House Committee previously noting the LLM's capabilities mirror those of American models.
OpenAI's submission marks the first public documentation of alleged IP circumvention by a Chinese AI firm. The company claims DeepSeek's methods constitute "free-riding on U.S. frontier lab research" through programmatic distillation.
