Microsoft is the lead investor in OpenAI and has long been the artificial intelligence startup’s key cloud partner. But in the latest sign that AI relationships are getting complicated, Microsoft is beginning to use more technology from OpenAI rival Anthropic.
The software giant said Wednesday that it’s starting to draw on an AI model from Anthropic to answer some queries in the Microsoft 365 Copilot assistant for commercial clients.
The effort represents another step toward diversification in generative AI for Microsoft, which has mainly relied on OpenAI models for artificial intelligence features in Bing, Windows and other products. As part of a strategic partnership, Microsoft has invested more than $13 billion in OpenAI, and OpenAI operates its models in the Azure cloud, among other places.
OpenAI has been busy expanding its roster of high-powered partners. In recent weeks, the company has disclosed that it plans to spend $300 billion with Oracle, and it’s forged a $10 billion agreement with chip designer Broadcom. This week, chipmaker Nvidia said it will invest up to $100 billion in OpenAI as part of a joint effort to spend hundreds of billions of dollars on new data centers.
Microsoft and OpenAI remain close, but their arrangement is changing.
Last year, Microsoft said it was allowing software engineers to get coding help from Anthropic and Google models in the GitHub Copilot Chat assistant, and not just from OpenAI. Anthropic, founded in 2021 by former OpenAI executives, was recently valued by investors at $183 billion. Microsoft remains the exclusive cloud supplier for OpenAI’s programming interface that third-party applications rely on.
Anthropic’s model is coming first to Researcher, an agent for Microsoft 365 Copilot that can gather and analyze information and generate reports.
Employees at organizations enrolled in Microsoft’s Frontier program, which provides early access to AI features, can now choose to use Anthropic’s Claude Opus 4.1 reasoning model, a type of model that’s designed to break down complex tasks, as an alternative to OpenAI’s competitive offering in Researcher.
“And stay tuned: Anthropic models will bring even more powerful experiences to Microsoft 365 Copilot,” Charles Lamanna, Microsoft’s president of business and industry Copilot, wrote in a blog post.
Corporate users can now opt to tap Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 when building their own AI agents with Microsoft’s Copilot Studio tool.
Administrators must enable Anthropic models before employees can use them. The models run in the Amazon and Google clouds. Their use is subject to Anthropic’s terms and conditions, Lamanna wrote.
In addition to incorporating large language models from more companies, Microsoft has started testing MAI-1-preview, which it built in-house.
“As AI becomes more capable and agentic, models themselves become more of a commodity,” Microsoft CEO Satya Nadella wrote in an October LinkedIn post.
Last year, Microsoft identified OpenAI and Anthropic as competitors in AI products.
The Information reported on Microsoft’s Copilot plans with Anthropic earlier in September.
Copilot is key to Microsoft’s AI strategy.
“It continues to be rolled out among enterprises in phases,” analysts at KeyBanc wrote in a July note to clients. The analysts, who recommend holding the stock, wrote that based on an “industry check,” about 60% of customers are “implementing Copilot for just 10% of their 365 user base,” and just 4% roll it out for every user.
