Microsoft (MS) announced that it will add the inference model ‘R1’ of Chinese AI startup DeepSeek to its cloud service Azure AI Foundry. MS stated that R1 has undergone thorough security and safety reviews, and that some lightweight versions will run on Copilot+, a Windows-based AI-supported PC, in the future. However, MS’s decision is causing controversy. According to Bloomberg, MS is working with OpenAI to investigate whether DeepSeek leaked a large amount of data without permission through OpenAI’s API in the fall of 2024, and suspects that DeepSeek may have stolen OpenAI’s intellectual property (IP). Nevertheless, as R1 is receiving a lot of attention in the AI industry, it is analyzed that MS has accepted the DeepSeek model to secure competitiveness. However, R1 has raised reliability and censorship issues, such as having an 83% error rate on news-related questions and not responding to 85% of China-related inquiries, so it is unclear whether MS has improved this.
MS Adds DeepSec ‘R1’ Model to Azure AI
