OpenAI claims to have found evidence of so-called "distillation," a technique that allows developers to improve the performance of smaller AI models by using the output from larger and more powerful models. Although distillation is a common practice in the AI industry, OpenAI claims that DeepSeek used this method to create its own model, which is against the company's rules. According to OpenAI's policy, users are prohibited from "copying" its services or using the results to develop competing AI models.
OpenAI did not provide specific evidence for its accusations, but emphasized that it actively monitors and blocks accounts suspected of violating the terms. OpenAI, along with key partner Microsoft, reportedly investigated and blocked several DeepSeek-related accounts last year over suspected use of distillation in violation of the company's rules. Microsoft declined to comment on the matter.
DeepSeek has not yet responded to the allegations, as China is celebrating the Chinese New Year.
Although OpenAI expresses concern about possible infringement of its intellectual property, some AI researchers argue that distillation is a common practice in both China and the United States. Ritvik Gupta, an AI researcher at the University of California, Berkeley, noted that many startups and academic institutions use the output from commercial AI models such as ChatGPT to improve their own systems.
"It is not surprising to me that DeepSeek supposedly would be doing the same. If they were, stopping this practice precisely may be difficult," Gupta explained.