Microsoft integrates Anthropic's Claude to critique OpenAI GPT drafts in Copilot Researcher agent

Microsoft announced a new feature for its Microsoft 365 Copilot Researcher agent on Monday. The capability, called Critique, directs OpenAI's GPT models to draft responses to research queries. Anthropic's Claude model then reviews the drafts for accuracy, completeness and citation quality. The company said it expects the process to run in both directions in the future, with Claude drafting and GPT critiquing. Microsoft reported that the multi-model approach has delivered a 13.8 percent improvement on the DRACO benchmark. The result places the system ahead of standalone deep-research tools from OpenAI, Google, Perplexity and Anthropic, the company said. The announcement forms part of Microsoft's push to increase adoption of its Copilot tools among business users. The company said in January that it had 15 million paid Copilot seats. That figure represents roughly 3.3 percent of its 450 million commercial Microsoft 365 users.