OpenAI accuses Chinese AI models of “distelling” its products and replicating its technology

According to Reuters, OpenAI has told American lawmakers that the Chinese artificial intelligence startup DeepSeek, which gained global attention a year earlier for developing its AI model for cheap, is attempting to replicate the capabilities of leading American AI models using unauthorized means.

In the document, OpenAI accused DeepSeek of “ongoing efforts to free ride on the capabilities developed by OpenAI and other U.S. frontier labs.”

The memo explains that the approach in question is known as distillation. This process uses a more advanced, established AI system to evaluate and refine a newer model’s responses, effectively transferring knowledge from the older system to the newer one.

OpenAI went further, accusing DeepSeek employees of unauthorized access. The company reportedly wrote: “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.”

Based in Hangzhou, DeepSeek made headlines in the financial markets early last year after unveiling a suite of AI models considered competitive with top offerings from U.S. companies and claiming that these products cost a fraction of the billions spent by OpenAI, Google, and other U.S.-based companies to develop LLMs.

DeepSeek’s advancement was so convincing last year that it caused a major sell-off of tech stocks in early 2025, including Nvidia shedding almost $600 billion of its market cap as market forces speculated that the rise of DeepSeek was a sign that the need for the company’s advanced chips might be overstated.

OpenAI said it actively monitors and removes users suspected of attempting to distill its systems to build competing models.