Microsoft investigates whether DeepSeek-affiliated group illegally obtained OpenAI data

Microsoft Corp. and OpenAI are investigating whether OpenAI technology data was unauthorizedly obtained by a group affiliated with Chinese artificial intelligence startup DeepSeek, Bloomberg reports, citing people familiar with the matter, UNN reports.

Details

Microsoft security researchers noticed in the fall that individuals they believed might be connected to DeepSeek were obtaining large amounts of data using the OpenAI application programming interface, or API, sources said. Software developers can pay for a license to use the API to integrate OpenAI's proprietary artificial intelligence models into their own programs.

Microsoft, OpenAI's technology partner and its largest investor, has notified OpenAI of the activity, the sources said. Such activity may violate OpenAI's terms of service or indicate that the group acted to remove OpenAI's restrictions on the amount of data they can obtain, the sources said.

Addendum

Earlier this month, DeepSeek released a new open-source artificial intelligence model called R1 that can mimic the way humans think, upending a market dominated by OpenAI and U.S. rivals such as Google and Meta Platforms Inc. The Chinese newcomer claimed that R1 competes with or outperforms products from leading US developers on a number of industry metrics, including math problems and general knowledge, and was created at a fraction of their cost. The potential threat to the advantage of US companies in the industry led to a drop in AI-related tech stocks on Monday, including Microsoft, Nvidia Corp., Oracle Corp. and Google's parent company Alphabet Inc. that resulted in a loss of nearly $1 trillion in market value.

David Sachs, US President Donald Trump's "artificial intelligence czar," said on Tuesday that there is "substantial evidence" that DeepSeek relied on the results of OpenAI models to help develop its own technology. In an interview with Fox News, Sachs described a method called distillation, in which one AI model uses the results of another to learn in order to develop similar capabilities.

"There's substantial evidence that DeepSeek has gained knowledge from OpenAI's models here, and I don't think OpenAI is very happy about that," Sachs said, without going into details.

In a statement responding to Sachs' comments, OpenAI did not comment directly on his comments about DeepSeek. "We are aware that companies from China - and others - are constantly trying to adopt the models of leading American AI companies," said an OpenAI spokesperson. - "As a leading developer of AI, we take countermeasures to protect our intellectual property, including a rigorous process of incorporating advanced capabilities into the models we release, and believe that as we move forward, it is critical to work closely with the U.S. government to best protect the most effective models from adversaries and competitors' attempts to take over U.S. technology.