Microsoft's close partner OpenAI has accused DeepSeek of stealing its intellectual property and violating service terms. Nevertheless, Microsoft still intends to use DeepSeek's new models on its cloud platform.
Today, Microsoft announced that DeepSeek's R1 model, referred to as a "reasoning model," has been introduced as part of the Azure AI Foundry service. Azure AI Foundry is a platform that combines various artificial intelligence services offered by Microsoft for businesses. According to Microsoft's blog post, the model has undergone security checks and automatic assessments to mitigate potential risks.
Microsoft stated that customers will soon be able to run simplified versions of the R1 model on Windows devices equipped with Copilot+ that meet certain AI preparation requirements. Additionally, Microsoft plans to continue increasing the number of models available on Azure AI Foundry, allowing developers and businesses to use the R1 model to solve real-world problems.
In the fall of 2024, Microsoft's security researchers discovered that DeepSeek extracted a significant amount of data using the OpenAI API. Given that Microsoft is OpenAI's largest shareholder, the company informed OpenAI about the issue. Despite this, the R1 model remains a focal point, and Microsoft is considering integrating it into its cloud services due to its continued interest. However, the accuracy of the R1 model's responses is still uncertain.
According to tests conducted by NewsGuard, a media reliability organization, the R1 model failed to provide answers or gave incorrect responses to news-related questions 83% of the time. In another test, the R1 model did not respond to 85% of queries related to China, which may be due to censorship in the country.
