Microsoft launches investigation against DeepSeek – and integrates R1 into Azure

The Chinese company DeepSeek, which is currently shaking up the markets with its AI models, is now also available in Microsoft's Azure AI Foundry and via GitHub. This allows Microsoft customers to integrate Model R1 into their services. This seems surprising, as Microsoft is also investigating together with OpenAI whether DeepSeek has violated its terms of use.

According to Microsoft, the Azure portfolio includes around 1800 AI models. “DeepSeek R1 is available on a trusted, scalable and enterprise-ready platform that enables organizations to seamlessly integrate advanced AI while meeting SLAs, security and responsible AI commitments – all backed by Microsoft's reliability and innovation.”

Subsequently, Microsoft even praises R1'scost-effectiveness, allowing more people to experience cutting-edge AI capabilities with minimal investment in infrastructure. This almost sounds like a dig at OpenAI. Until now, Microsoft was a major investor in OpenAI and the AI company was bound to use the backer's infrastructure. This contract was recently relaxed. OpenAI is now also allowed to access other cloud providers. This is likely to affect the construction of a gigantic data center as part of the Stargates project, for example, which is intended to benefit OpenAI.

Azure AI Foundry offers users the opportunity to test and compare AI models directly. This is primarily aimed at business customers. DeepSeek R1 underwent numerous security tests before integration, says Microsoft.

Investigations into DeepSeek

Nevertheless, Microsoft, like OpenAI, is said to have launched an investigation against DeepSeek. The allegation is that the terms of use have been breached. Specifically, DeepSeek is alleged to have accessed data from OpenAI in an unauthorized manner. They are said to have used the API to do so. According to Bloomberg, Microsoft detected suspicious activity in the fall. OpenAI's terms of use state that it is forbidden to use the output to develop your own models that compete with those of OpenAI. This is exactly what DeepSeek is said to have done.

Several AI experts had expressed suspicions early on that DeepSeek had used model distillation, a process for transferring the knowledge of a large model to a smaller one. This is supported by the fact that DeepSeek's models are said to have replied several times that they are ChatGPT. Other answers also led to the conclusion – for example, information on the limitation of knowledge.

DeepSeek's models also come with their restrictions, namely those that suit the Chinese government. There are no answers to questions such as what happened in Tiananmen Square. Protests by a democracy movement were bloodily ended there in 1989.

(emw)