OpenAI Accuses DeepSeek of Using ChatGPT to Train Rival Chatbot

In today’s installment of Irony Daily, OpenAI is not happy someone stole its data to train an AI chatbot. The San Francisco-based tech giant claims Chinese AI startup DeepSeek used ChatGPT and its proprietary AI models to train competitors — reports the Financial Times.

OpenAI told the publication it has evidence of “distillation,” allegedly from DeepSeek. Distillation is when smaller AI models are trained on outputs from larger, better models to up the quality of their responses and cut down training costs. It is a common practice among AI labs across the globe, but doing it to create a competing chatbot violates OpenAI’s terms of service.

“The issue is when you [take it out of the platform and] are doing it to create your own model for your own purposes,” said one person close to OpenAI. The company, which has been battling copyright infringement claims from newspapers and media publications over training its models on their content without consent, is miffed at the potential theft of its intellectual property.

OpenAI and its biggest backer, Microsoft, reportedly investigated and blocked accounts allegedly belonging to DeepSeek last year for using OpenAI’s API for distillation. OpenAI has yet to provide details on the evidence it claims to have against DeepSeek.

DeepSeek this month released its open-source AI models, supposedly developed at a fraction of the cost incumbents like OpenAI, Google, and Meta have put into their offerings. The company’s AI chatbot, powered by the DeepSeek-V3 model, recently overtook ChatGPT as the #1 free app on Apple’s App Store.

DeepSeek’s meteoric rise and the solid performance of its models have shaken investor confidence in leading Western AI giants and brought into question what it really takes to develop a leading AI model.

AI experts found the responses from DeepSeek’s chatbot to be on par with, if not better than, rivals like ChatGPT. However, some said its responses indicated it may have been trained on outputs from OpenAI’s GPT-4 model.

U.S. President Donald Trump’s AI and crypto tsar, David Sacks, said during a segment on Fox News that “it is possible” DeepSeek stole OpenAI’s intellectual property. “There’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI models, and I don’t think OpenAI is very happy about this,” he added.

OpenAI has expressed concerns over Chinese companies copying U.S. technology. The company said in a statement that it wants to work closely with the U.S. government to safeguard the best AI models from overseas competitors.