That report from a few weeks ago was spot-on. As The Information’s sources tipped, Microsoft 365 Copilot is adding Anthropic’s AI models. Microsoft announced today that Claude access is now rolling out for beta testers. For starters, it’s being integrated into Copilot’s Researcher and its agent development tool.
“Copilot will continue to be powered by OpenAI’s latest models,” Microsoft’s announcement was quick to stress. Following that report from earlier this month, it was easy to imagine growing tensions between the two. It didn’t help that the story came on the heels of reports of uneasy negotiations between the pair.
Whether or not there’s anything to that, OpenAI models like GPT-5 still fuel most of Copilot… for now. But Microsoft 365’s Researcher now allows testers to use Anthropic’s Claude Opus 4.1. Copilot’s reasoning tool launched earlier this year. It’s designed to help you “tackle complex, multistep research.”
Microsoft posted the brief video below. It shows a new “Try Claude” button in the upper-right corner of the research agent’s interface.
Copilot Studio, Microsoft’s AI agent dev tool, offers similar functionality. There, a drop-down menu lets you choose between OpenAI’s and Anthropic’s models.
Anthropic integration is currently rolling out through Microsoft’s Frontier program. (That’s its optional beta program for AI previews.) After signing up, you’ll need to opt in to use Anthropic. This will all presumably roll out to the wider public later.
This won’t be the end of Copilot’s Anthropic integration. “Anthropic models will bring even more powerful experiences to Microsoft 365 Copilot,” Microsoft’s Charles Lamanna teased.
