Microsoft adds Anthropic AI to Copilot

Microsoft's Copilot is generally based on AI models from its partner OpenAI. Because this is not God-given, Microsoft is now gradually introducing alternatives: customers will be able to switch to Anthropic's AI models if and when they want to. At its subsidiary Github, Anthropic is already the first choice in Visual Code Studio for paying subscribers to Github Copilot.

It is now starting in Microsoft Copilot Studio, the application with which AI agents can be created. After an opt-in, the Anthropic models Claude Sonnet 4 and Claude Opus 4.1 can be activated. Anyone using the pre-programmed agent called Researcher as part of the Frontier program can also switch this to Claude Opus 4.1. Frontier is intended for customers who always prefer the latest technology and are willing to accept a higher risk in return.

Microsoft 365 Copilot follows

The changeover is not monolithic but is intended to be flexible so that different tasks can be handled with different AI models. However, if an anthropic model is selected, the data will not remain in Microsoft's cloud but will move to Amazon.com's AWS, which hosts the anthropic models. Of course, it doesn't have to stay that way; if the offer is attractive, Microsoft will certainly be happy to make Anthropic a hosting offer.

“This is just the beginning,” writes Microsoft President Charles Lamanna, responsible for Business & Industry Copilot, in a blog post. “We are committed to delivering AI model innovation quickly.” The goal is to transform every business process with AI agents. “And stay tuned: Anthropic models will deliver even more powerful experiences with Microsoft 365 Copilot.” The data company is not revealing a timetable for this.

By the way, Anthropic is changing its terms of use on 28 September. From then on, chat transcripts and code sessions from the consumer versions of Claude will be analyzed for the training of new AI models, unless the user has objected.

(ds)