Full Breakdown
OpenAI Accuses DeepSeek of Distilling U.S. AI Models
2/13/2026, 10:52:24 AM
Allegations of Intellectual Property Misuse
OpenAI has formally accused the Chinese artificial intelligence startup DeepSeek of employing sophisticated methods to extract outputs from its AI models, aiming to train competing systems. This accusation was detailed in a memo submitted to the U.S. House Select Committee on Strategic Competition between the United States and the Chinese Communist Party. OpenAI claims that DeepSeek has been using a technique known as "distillation," which allows a newer AI model to learn from the outputs of a more advanced model, effectively transferring knowledge without direct access to proprietary data.
Techniques and Methods Identified
The memo indicates that DeepSeek employees have developed methods to circumvent OpenAI's access restrictions, utilizing obfuscated third-party routing mechanisms to mask their origins while accessing U.S. AI systems. OpenAI reported that these employees created code designed to programmatically collect large volumes of model outputs for training purposes. This practice raises significant concerns regarding the unauthorized use of U.S. technology and the potential erosion of American leadership in AI.
Context and Implications
DeepSeek, based in Hangzhou, gained attention last year for releasing AI models that rival those of leading U.S. companies, intensifying fears in Washington that Chinese firms could close the technological gap despite existing export controls on advanced chips. OpenAI has expressed concerns that Chinese developers are cutting corners on safety practices during the training and deployment of their models. The memo highlights that the distillation process, while common in research, poses risks when applied without authorization, particularly in sensitive areas like biology and chemistry.
Official Responses and Statements
Representative John Moolenaar, chair of the House China committee, remarked that the situation exemplifies the Chinese Communist Party's strategy of "steal, copy, and kill," emphasizing the ongoing threat posed by DeepSeek's actions. OpenAI has stated that it routinely removes users suspected of attempting to extract model outputs for competitive purposes and is enhancing its monitoring and enforcement mechanisms to prevent unauthorized access.
Criticism and Concerns
Critics have raised alarms about the implications of DeepSeek's advancements, particularly regarding national security. White House AI Czar David Sacks previously warned about the risks associated with Chinese distillation tactics, suggesting that DeepSeek is leveraging older technology to gain an edge. Additionally, Republican Representative Michael McCaul expressed concern over the potential consequences of providing advanced semiconductor chips to China, fearing that such actions could further empower Chinese AI capabilities.
Conflicting Reports & Gaps
While OpenAI has detailed its concerns and the methods employed by DeepSeek, responses from DeepSeek and the Chinese embassy in Washington have not been forthcoming. This lack of comment leaves a gap in understanding the full context of the allegations and the company's perspective.
Verbatim Quotes
- “OpenAI has observed ongoing efforts to free-ride on the capabilities developed by OpenAI and other U.S. frontier labs,” — OpenAI Memo
- “this is part of the CCP’s playbook: steal, copy, and kill,” — Representative John Moolenaar
- “DeepSeek should have been a wake-up call about the dangers of selling advanced semiconductor chips to the CCP.” — Representative Michael McCaul
The situation underscores the growing tensions in the AI sector, as U.S. companies grapple with the implications of foreign competition and the potential for intellectual property theft.
