OpenAI and Microsoft are investigating whether Chinese artificial intelligence company DeepSeek leveraged OpenAI’s models to train its own AI, according to a Bloomberg report. The inquiry comes after DeepSeek, known for its cost-effective AI development, introduced models that compete with OpenAI’s flagship offerings, triggering concerns about potential intellectual property violations.
Sources cited in the report suggest that Microsoft security teams detected unusual data movements in late 2024, involving large-scale data extraction through OpenAI developer accounts allegedly linked to DeepSeek. OpenAI also told the Financial Times that it found evidence of AI model distillation, a process where smaller AI models are trained using data extracted from more powerful models.
While distillation is a common technique in AI development, using OpenAI’s API to systematically extract and replicate its models would violate OpenAI’s terms of service. OpenAI has not publicly shared the specifics of the evidence but maintains that DeepSeek may have used this approach to build its competitive models at a fraction of the $100 million+ spent on training GPT-4.
The controversy has caught the attention of President Donald Trump’s administration. White House AI policy advisor David Sacks told Fox News that the allegations could indicate intellectual property theft. “There’s substantial evidence that what DeepSeek did here is they distilled knowledge out of OpenAI models, and I don’t think OpenAI is very happy about this,” Sacks said.
A statement from OpenAI to Bloomberg reinforced concerns about foreign attempts to exploit US-developed AI. “We know PRC (China)-based companies—and others—are constantly trying to distill the models of leading US AI companies,” OpenAI said. The company added that it is working on countermeasures to protect its intellectual property and is collaborating with the US government to prevent foreign entities from leveraging American AI advancements.
While OpenAI and Microsoft continue their investigation, DeepSeek has not publicly responded to the allegations.