Microsoft and OpenAI are looking into whether a group connected to the Chinese AI startup DeepSeek gained unauthorized access to data from OpenAI’s technology.
Sources say that last fall, Microsoft’s security team detected what they believe were attempts by individuals linked to DeepSeek to exfiltrate large amounts of data through OpenAI’s API, according to Bloomberg.
The API allows developers to pay for licenses to integrate OpenAI’s AI models into their own applications.
Microsoft, which is both a key partner and the largest investor in OpenAI, alerted the company about the suspicious activity, according to sources.
This behavior could potentially breach OpenAI’s terms of service or suggest that the group attempted to bypass OpenAI’s restrictions on the amount of data they could access.
Earlier this month, DeepSeek unveiled R1, an open-source AI model designed to mimic human reasoning.
The launch challenges the dominance of OpenAI and U.S. competitors such as Google and Meta Platforms Inc.
DeepSeek claims that R1 rivals or even outperforms leading U.S. models across several industry benchmarks, including math and general knowledge, all while being built at a fraction of the cost.
The threat posed by DeepSeek’s R1 model to the U.S. companies’ dominance in the AI sector caused a sharp drop in technology stocks on Monday. Companies tied to AI, such as Microsoft, Nvidia, Oracle, and Google parent Alphabet, saw their market values collectively shrink by nearly $1 trillion.
