Microsoft has announced the launch of DeepSeek’s reasoning model, R1, on its Azure AI Foundry service, which consolidates various AI tools for businesses. This move comes despite concerns from Microsoft’s partner, OpenAI, suggesting that DeepSeek may have misused its intellectual property and violated terms of service.
In a blog post, Microsoft emphasized that the R1 model available on Azure has undergone extensive safety evaluations, including automated assessments and thorough security reviews to reduce potential risks.
In the near future, customers will also be able to use “distilled” versions of R1 on Copilot+ PCs, which are designed to meet specific AI readiness criteria. Microsoft expressed excitement about expanding its model catalog on Azure AI Foundry, highlighting the potential for developers and businesses to use R1 for real-world applications and transformative solutions.
The decision to include R1 in Microsoft’s cloud offerings is intriguing, especially since the company is reportedly investigating DeepSeek for potential misuse of both its and OpenAI’s services. Security researchers at Microsoft have indicated that DeepSeek might have extracted significant data through OpenAI’s API in late 2024, leading Microsoft, a major shareholder of OpenAI, to inform them of these concerns.
Despite these issues, R1 remains a focal point of interest, possibly influencing Microsoft’s choice to integrate it into their cloud services.
However, it is still uncertain if Microsoft has made any changes to R1 to enhance its performance or address censorship issues. Tests conducted by NewsGuard revealed that R1 provides incorrect or no answers 83% of the time on news topics and declines to respond to 85% of inquiries related to China, likely due to the censorship imposed on AI models in that region.
