OpenAI has alleged that DeepSeek engaged in malpractice while developing the next version of its AI model, even prior to its official release.

“DeepSeek’s next model (whatever its form) should be understood in the context of its ongoing efforts to free-ride on the capabilities developed by OpenAI and other US frontier labs,” OpenAI said in a memo to the U.S. House Select Committee on China on February 12.

DeepSeek has not confirmed any new releases.

OpenAI’s concerns appear linked to expectations that the Hangzhou-based company might make a major announcement during next week’s Lunar New Year celebrations, reminiscent of the surprise launch it unveiled last year.

DeepSeek shot to the forefront of the global AI race almost overnight with the release of its R1 model during last year’s Lunar New Year, claiming performance on par with leading U.S. models despite being trained using far fewer advanced chips.

The launch reignited discussions in Washington over whether U.S. export controls on high-end semiconductors are enough to maintain America’s AI advantage.

In its memo, OpenAI accused DeepSeek of employing “distillation” techniques—a standard approach in which a smaller model is trained on the outputs of a larger, more powerful model to mimic its capabilities.

“We have observed accounts associated with DeepSeek employees developing methods to circumvent OpenAI’s access restrictions and access models through obfuscated third-party routers and other ways that mask their source,” the memo said. “We also know that DeepSeek employees developed code to access U.S. AI models and obtain outputs for distillation in programmatic ways. We believe that DeepSeek also uses third-party routers to access frontier models from other U.S. labs.”

OpenAI stated that it does not allow its outputs to be used to develop “imitation frontier AI models” that replicate its capabilities.

This is not the first time OpenAI has flagged concerns over distillation; the company investigated whether DeepSeek had used its data in this way shortly after the R1 model launched in January last year.

DeepSeek’s R1 model has fueled China’s adoption of open-weight AI models—systems that developers globally can download, modify, and deploy. This approach contrasts with the closed systems preferred by most U.S. tech giants, which tightly restrict access to their models, data, and architecture.