Microsoft is currently investigating claims that data from OpenAI’s models might have been accessed and exploited without authorization by a group linked to the Chinese artificial intelligence firm DeepSeek.
According to a report by Bloomberg on Wednesday, Microsoft’s security team reportedly identifies individuals believed to be related to DeepSeek extracting a significant amount of data from OpenAI’s application programming interface, or API. The API is an interface through which paid software developers- including Microsoft-receive access to OpenAI proprietary AI models.
Microsoft is one of the largest investors in OpenAI and has its AI technology integrated into all of Microsoft’s products. The probe is timely as DeepSeek continues to face rising criticisms in various spheres following the release of its AI model, R1.
The R1 model, which appeared to perform similarly to OpenAI’s ChatGPT, quickly became the most downloaded free app on Apple’s App Store. What was particularly striking, however, was the model’s ability to achieve such high performance using less advanced hardware and a fraction of the budget that larger competitors, like OpenAI, typically invest in their systems.
Sources familiar with the probe reportedly told Bloomberg that Microsoft researchers were investigating the extent to which DeepSeek might have accessed data hosted on OpenAI. DeepSeek, of course, is reportedly a firm back by Chinese hedge fund High-Flyer; it is prompting questions of whether the company, through some unauthorized access, could have illicitly utilized proprietary data drawn from the most prominent AI platforms-including ChatGPT-to create the model.
DeepSeek’s surprise surge has thrown the world markets into a tizzy, especially in the semiconductor space. The stocks of chipmakers, which are one of the primary players in AI infrastructure, plummeted after it entered the market. The company’s release of the R1 model has raised questions about the competitive landscape of AI, with concerns over intellectual property potentially being exploited.
Meanwhile, David Sacks, former Trump administration’s president of the initiative on AI, also made comment on the development. Sacks said that DeepSeek probably engaged in a practice called “distillation,” where one AI model takes the output of another to train a new system with identical capabilities. Sacks referenced the “substantial evidence” he saw of this practice, which only provided further reason to think DeepSeek merely used the data from OpenAI for its benefit.
