Microsoft brings Deepseek to its cloud

The close and collaborative Microsoft partner, Openai, may suggest that Deepseek Stole his IP address and violate his service conditions. But Microsoft still wants the new shiny Deepsek models on its cloud platform.

Microsoft announced today R1The so -called alleged thinking model is available on Azure ai Mabka The service, the Microsoft platform that collects a number of AI services for institutions under one banner. in Blog postMicrosoft said that the R1 version on Azure Ai Foundry “has undergone collective designs and strict red safety”, including “automatic assessments of model behavior and extensive security reviews to reduce potential risks.”

In the near future, Microsoft said, customers will be able to use “distilled” flavors from R1 to operate locally Coilot+ pcsMicrosoft brand of Windows devices that meet some of the requirements for artificial intelligence.

“With the continued expansion of the model catalog in the Azure AI, we are excited to learn how developers and institutions (…) R1 benefit to address the challenges in the real world and provide transformative experiences,” Microsoft continued in the post.

The R1 addition to Microsoft Cloud Services is a strange offer, taking into account that Microsoft According to what was reported, the investigation started In the possible use of Deepseek for its services and Openai. According to security researchers working for Microsoft, Deepseek may have clarified a large amount of data using API from Openai in the fall of 2024. Microsoft, which also coincides with the largest contributor to Openai, Openai was notified of suspicious activity, For each Bloomberg.

But R1 is talking about the city, and Microsoft has been persuaded to bring it to its cloud while it is still bearing gravity.

It is unclear whether Microsoft has made any modifications to the model to improve its accuracy – combating censorship. According to the test Through the Newsguard-R1 provides inaccurate or non-answers to 83 % of time when asked about news related topics. Separate test I found that R1 refuses to answer 85 % of China’s claims Government control that is subject to artificial intelligence models in the country.