In response, Microsoft has officially prohibited its employees from using the DeepSeek application. This decision comes in response to legitimate national security and data privacy concerns, notably foreign influence. The reasoning for the ban is spelled out by the company’s president, Brad Smith, who wrote about it here, arguing for the need to protect confidential data.
DeepSeek’s R1 model, which is accessible via Microsoft’s Azure cloud service, gained quick traction shortly after its release. Since this is an open-source model that anyone can download and run on their own servers, it has sparked concerns over data storage and security. The model’s architecture permits users to offer services to clients without transmitting data back to China, yet Microsoft remains wary of its implications.
Those concerns include DeepSeek’s nakedly cavalier privacy policy that states user data is kept on servers in the PRC. This situation is made worse by the model’s reported propensity to censor topics considered sensitive by the Chinese government. As Microsoft claimed at the time, these types of censorship and data storage practices might leave users open to “Chinese propaganda.”
The stage for Microsoft’s dramatic play is the backdrop of technology-first rivalry. Earlier this month, the Biden administration placed sweeping sanctions on four Chinese technology entities. In the case of China’s largest artificial intelligence company, investigators are examining the company’s practices and impact.
“At Microsoft we don’t allow our employees to use the DeepSeek app.” – Brad Smith
DeepSeek has even become a more recent direct competitor to Microsoft’s own newly released in-house Copilot internet search chat app. Despite its competitive positioning, Microsoft has not banned all chat applications from its Windows app store; for instance, Perplexity remains available for users. DeepSeek’s connection to Chinese law—which mandates the product’s compliance with the country’s intelligence arms—adds another layer of complexity to its position.
Microsoft used to place DeepSeek through “intensive red teaming and safety assessments” before offering it up on Azure. The company soon discovered that these efforts weren’t enough to address the larger questions of data privacy. They understood the power of propaganda to influence public opinion.
Microsoft had previously subjected DeepSeek to “rigorous red teaming and safety evaluations” prior to making it available on Azure. However, the company determined that these measures were insufficient in addressing broader concerns regarding data privacy and the potential for propaganda influence.
