ChatGPT Maker Suspects China’s Dirt Cheap DeepSeek AI Models Were Built Using OpenAI Data — and the Irony Is Not Lost on the Internet

OpenAI has voiced concerns that China's DeepSeek AI models, known for their low cost, may have been developed using OpenAI's data. This revelation, coupled with DeepSeek's market impact, has prompted Donald Trump to call it a wake-up call for the U.S. tech industry. Nvidia, a major player in the GPU market crucial for AI, experienced a significant stock drop of 16.86%, the largest in Wall Street history, following DeepSeek's emergence. Other tech giants like Microsoft, Meta, Alphabet, and Dell also saw their stock prices decline.

DeepSeek's R1 model, built on the open-source DeepSeek-V3, is marketed as a significantly cheaper alternative to Western AI models like ChatGPT. Its reported training cost of just $6 million, although disputed by some, has raised questions about the massive investments made by American tech companies in AI, unsettling investors. DeepSeek's popularity, evidenced by its top ranking on U.S. app download charts, further fuels this discussion.

Bloomberg reports that OpenAI and Microsoft are investigating whether DeepSeek utilized OpenAI's API to integrate OpenAI's AI models into its own. OpenAI acknowledges that Chinese companies, and others, actively attempt to leverage leading U.S. AI companies' models through techniques like distillation—a process of extracting data from larger models to train smaller ones, which violates OpenAI's terms of service. OpenAI emphasizes its commitment to protecting its intellectual property and collaborating with the U.S. government to safeguard its technology.

David Sacks, President Trump's AI czar, suggests evidence points to DeepSeek using OpenAI models through distillation. He anticipates leading AI companies will implement measures to prevent such practices. This situation highlights the irony of OpenAI's position, given previous accusations of its own data acquisition practices. Critics point to OpenAI's past statements, including a 2024 assertion that creating AI tools like ChatGPT without copyrighted material is impossible, and its submission to the UK's House of Lords emphasizing the reliance on copyrighted work for training large language models. These statements contrast sharply with OpenAI's current concerns about DeepSeek's actions. The ongoing legal battles, including lawsuits from the New York Times and 17 authors, further underscore the complexities and controversies surrounding the use of copyrighted material in AI model training. The 2018 U.S. Copyright Office finding that AI art cannot be copyrighted adds another layer to this evolving legal and ethical landscape.