OpenAI accuses DeepSeek of unlawfully copying ChatGPT technology

OpenAI has accused Chinese AI lab DeepSeek of using its technology to train the new R1 model.

According to Bloomberg, DeepSeek relies on a technique known as distillation, extracting outputs from U.S. AI models and using them to build comparable capabilities. OpenAI argues that the practice poses not only a commercial threat but also potential national security risks for the United States.

The company claims that distillation can bypass model safeguards and enable high-risk applications in areas such as biology or chemistry. OpenAI alleges that DeepSeek accessed model outputs through indirect channels to circumvent usage restrictions. In response, U.S. authorities have launched an investigation and tightened export controls on advanced AI chips to China in an effort to limit the spread of American technology.

Earlier in 2025, DeepSeek drew global attention after releasing a series of models that challenged assumptions about the cost and efficiency of AI development. Its DeepSeek-R1 model, in particular, demonstrated reasoning capabilities comparable to leading Western systems such as OpenAI’s o1, reportedly at a fraction of the training cost.