Microsoft Copilot now uses GPT and Claude together for more accurate answers

Microsoft has rolled out two new features in Copilot that signal a strategic shift in enterprise AI: Critique and Council. Both became available to early access users through the Frontier program in late March and early April 2026, and are now rolling out to all eligible customers.

Critique works by splitting writing and evaluation between two different AI models. Typically, GPT drafts a research report while Anthropic's Claude acts as an expert reviewer. The reviewer checks the output for accuracy, source quality, and completeness, with the goal of reducing hallucinations and building trust in the final answer.

Council takes a different approach. Two models, one from OpenAI and one from Anthropic, run in parallel and produce two independent reports on the same topic. A third judge model then distills a summary that highlights where the two agree and where they diverge in framing or interpretation. This is particularly useful in high-stakes decision scenarios where multiple perspectives matter.

For enterprise teams already running Microsoft 365, these features are built directly into the Researcher agent with no additional configuration. Research tasks, competitive analysis, and due diligence workflows can benefit immediately.

Microsoft's broader message is clear: the company is not betting on a single AI provider. Critique is set as the default in auto mode, while Council is available as an explicit choice. The customer decides.