Microsoft has announced that DeepSeek’s R1 model is now available on its Azure AI Foundry platform, even as the company investigates DeepSeek for potential intellectual property violations related to OpenAI.
This move brings R1, a model focused on reasoning capabilities, to Microsoft’s cloud services, allowing enterprises to integrate it into their AI workflows.
In a blog post, Microsoft stated that R1 has undergone extensive safety evaluations, including red teaming and automated behavior assessments. The company also revealed plans to offer optimized versions of R1 for Copilot+ PCs, a line of AI-ready Windows devices.
However, the decision to onboard R1 is controversial. Reports suggest that Microsoft security researchers flagged DeepSeek last year for allegedly extracting large amounts of data using OpenAI’s API, leading Microsoft to alert OpenAI about possible misuse. Despite these concerns, the popularity of R1 appears to have influenced Microsoft’s decision to include it in its AI ecosystem.
It remains unclear whether Microsoft has modified R1 to improve accuracy or address censorship issues. Tests by NewsGuard found that R1 provides inaccurate or non-answers 83% of the time on news-related queries and refuses to respond to 85% of questions about China—potentially reflecting government-imposed restrictions on AI models developed in the country.
