The close partner and collaborator of Microsoft, Openai, may suggest that Deepseek has stolen his IP and violated his terms of service. But Microsoft still wants the new fantastic Deepseek models on its cloud platform.
Microsoft has announced today that R1, the so -called Deepseek reasoning model, is available on the Azure Ai foundry service, the Microsoft platform that brings together a series of artificial intelligence services for companies under a single banner. In a blog post, Microsoft said that the version of R1 on Azure to the Foundry has “immediately rigorous red team assessments and safety”, including “automated evaluations of model behavior and large safety reviews to mitigate potential risks”.
In the near future, Microsoft said, customers will be able to use R1’s “distilled” flavors to work locally on Copilot+ PCS, the Microsoft Windows hardware brand that meets certain promptness requirements of the AI.
“While we continue to expand the model catalog in Azure to the Foundry, we are excited to see how developers and companies exploit (…) R1 to face the challenges of the real world and offer transformative experiences,” continued Microsoft in the post.
The addition of R1 to Microsoft’s cloud services is curious, considering that Microsoft has started a probe on the potential abuse of Deepseek of its Openni services. According to security researchers who work for Microsoft, Deepseek may have exfilt a large amount of data using Openi’s bees in the autumn of 2024 Suspected, for Bloomberg.
But R1 is the speech of the city and Microsoft may have been persuaded to bring him to his cloud turn while still keeps charm.
It is not clear if Microsoft has made changes to the model to improve its precision and fight its censorship. According to a test by the organization of reliability of newsguard information, R1 provides imprecise or non -answers 83% of the time when the news has been asked for the news. A separate test found that R1 refuses to respond to 85% of China instructions, probably a consequence of the government censorship to which models to developed in the country are subject.
Techcrunch has a newsletter focused on artificial intelligence! Sign up here to take it to your mailbox every Wednesday.
