OpenAI has cautioned U.S. lawmakers that Chinese artificial intelligence startup DeepSeek is allegedly attempting to replicate advanced American AI systems and use them to train its own models. According to a memo reviewed by Reuters, the ChatGPT developer claims that DeepSeek is targeting both OpenAI and other leading U.S. AI companies to extract technological advantages.
The company, led by Sam Altman, accused DeepSeek of engaging in “ongoing efforts” to benefit from capabilities developed by OpenAI and other frontier American AI labs. The concern centers on a method known as distillation, a process in which a more advanced AI model evaluates and refines the outputs of a newer system, effectively transferring knowledge from one model to another.
In a letter sent Thursday to the U.S. House Select Committee on Strategic Competition between the U.S. and the Chinese Communist Party, OpenAI stated that it had identified accounts linked to DeepSeek employees attempting to bypass access restrictions. The memo alleges that these efforts included routing requests through third-party systems designed to conceal their origin.
OpenAI further claimed that DeepSeek staff created code to programmatically access U.S. AI models and collect outputs for distillation purposes. DeepSeek and its parent company, High-Flyer, did not immediately respond to requests for comment.
Based in Hangzhou, DeepSeek drew significant attention last year after launching AI models that rivaled top U.S. offerings. The rapid progress intensified concerns in Washington that China could narrow the artificial intelligence gap despite export controls and regulatory restrictions.
OpenAI also warned that some Chinese large language models may be accelerating development at the expense of safety standards when training and deploying new systems. However, several Silicon Valley executives have previously praised DeepSeek’s globally available models, including DeepSeek-V3 and DeepSeek-R1.
The U.S. AI company added that it actively monitors and removes accounts suspected of attempting to distill its proprietary models to build competing technologies.
