Microsoft, OpenAI probe possible data breach linked to Chinese startup DeepSeek

Microsoft’s security team detected unusual activity last fall, with individuals suspected of being linked to DeepSeek extracting significant amounts of data through OpenAI’s API, Bloomberg reported on Tuesday, citing sources familiar with the situation.

This API, available to licensed developers, allows them to integrate OpenAI's AI models into their own applications.

As a major investor in OpenAI, Microsoft informed the company about the breach, which could potentially violate OpenAI's terms of service or indicate an effort to circumvent data access restrictions, the broadcaster said.

This investigation coincides with DeepSeek's rise in the AI industry, following the release of its open-source model, R1, which the company claims rivals or outperforms leading US AI products from OpenAI, Google, and Meta.

In an interview with Fox News on Tuesday, President Trump’s “AI czar” David Sacks said, “There’s substantial evidence that DeepSeek distilled knowledge out of OpenAI models to develop their own technology,” referring to the practice where AI models use the outputs of other systems for training.

Earlier this month, DeepSeek released its open-source AI model, R1, which mimics human reasoning and challenges the dominance of OpenAI and US competitors like Google and Meta.

The Chinese startup asserts that R1 competes with or surpasses US products in areas such as mathematics and general knowledge while being much cheaper to develop. This perceived threat to US companies led to a sharp decline in AI-related tech stocks, erasing almost $1 trillion in market value.