Microsoft first raises doubts about DeepSeek and then adds it to its cloud

Despite initiating a probe into Chinese AI startup DeepSeek, Microsoft has added the startup’s latest reasoning model R1, to its model catalog on Azure AI Foundry and GitHub.

Azure AI Foundry, which is a rebranded version of Azure AI Studio, comprises the Azure AI Foundry portal, which was earlier the Azure AI Studio, the Azure AI Foundry software development kit (SDK), Azure AI Agents, and pre-built app templates along with some tools for AI-based application development.

[ Related: More DeepSeek news and analysis ]

As part of the blog post in which Microsoft declared that it has added R1, it said that the model had “undergone rigorous red teaming and safety evaluations, including automated assessments of model behavior and extensive security reviews to mitigate potential risks.”

DeepSeek’s R1 is seen as a strong competitor to OpenAI’s o1.

The Microsoft blog noted that its customers would soon be able to use distilled flavors of the R1 model to run locally on their Copilot+ PCs.

While accessing the R1 model via Azure AI Foundry and GitHub, enterprises can try, compare, and implement this model in their code for free in the playground or via the API.

However, despite adding the R1 model to its cloud, Microsoft and OpenAI are reportedly probing the startup that developed the R1 model for IP infringement.

A separate report from Bloomberg noted that both the US-headquartered firms were checking if DeepSeek had accessed and copied OpenAI’s technology while developing its reasoning model.

The report further cites Microsoft security researchers saying that they had observed people affiliated or connected to DeepSeek taking out large amounts of data using OpenAI’s API — the tool that is used to interact with any OpenAI services or models.

Microsoft, which is also an investor in OpenAI, reportedly notified OpenAI of the alleged suspicious activity.

Another reason for the probe is the rising popularity of the DeepSeek model among users and cloud providers. The day it launched, it soared over ChatGPT — OpenAI’s assistant, in terms of downloads and rank on the Apple App Store.

The White House also seems to believe that DeepSeek may have found a way to copy some technology from OpenAI to create its R1 model.

In an interview to Fox News, David Sacks, often known as the White House’s AI czar, alluded that DeepSeek possibly infringed on OpenAI’s technology.

“There’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI’s models,” Sacks was quoted as saying.

Additionally, Reuters reported that an OpenAI spokesperson, when reached for comment, said that Chinese companies were always trying to replicate models coming out of leading US model providers.

Separately, White House and European regulators are investigating DeepSeek over national security risks, data privacy concerns, and potential IP violations. While the White House confirmed on Tuesday that the National Security Council (NSC) is reviewing the AI model’s implications as fears mount that Chinese advancements in AI could threaten the dominance of US-based AI firms including OpenAI and Google, Italy’s data protection authority, the Garante, launched its own investigation into DeepSeek, demanding clarity on its data collection practices.