OpenAI Calls for Ban on Chinas DeepSeek AI Models in the U.S.

The OpenAI company has recommended that the U.S. government prohibit AI models developed by the Chinese laboratory DeepSeek, citing concerns that the project is “state-subsidized” and “state-controlled.”

This recommendation is part of the initiative put forth by the Trump administration known as the «AI Action Plan.» The document claims that products from DeepSeek, including the notorious R1 neural network, are unsafe as they “comply with Chinese government demands for user data.”

A ban on using models from China across all countries classified as “level one” allies of the U.S.—which, according to regulations set forth by the Biden administration, have nearly unrestricted access to American AI technologies and semiconductors—would mitigate risks related to privacy and security, including “the threat of intellectual property theft,” claims the American startup.

It remains unclear whether the call to restrict pertains to DeepSeek’s open models, which do not include mechanisms for collecting and transmitting user data to the Chinese government. Companies such as Microsoft, Perplexity, and Amazon host these models on their infrastructure.

Earlier, OpenAI suspected that DeepSeek may have engaged in data theft. Nevertheless, CEO Sam Altman described DeepSeek-R1 as “impressive,” given its capabilities for the price.

The startup has also urged the U.S. government to develop an intellectual property strategy that will “maintain the ability of American models to learn from copyrighted materials.”

“There are so many AI startups, investments, and scientific breakthroughs in America largely due to the doctrine of fair use fostering the growth of artificial intelligence,” OpenAI writes.

As a reminder, in August 2024, the company announced a partnership with Condé Nast to display content from major publications.