Microsoft released Critique and Council, two new features that turn its Copilot Researcher into a multi-model system that can review and edit research reports and run both systems side by side to see where they agree and disagree.
Here are the details:
- Copilot’s Researcher already uses OAI for multi-step work, with Critique now adding Claude as a second model to review every report before it ships.
- One model drafts the research, and the second tears it apart on source quality, completeness, and evidence grounding behind the scenes.
- A separate Model Council mode runs both models side by side, then flags where they agree, where they split, and what each uniquely surfaced.
- The updates come alongside a broader rollout of Copilot Cowork into Frontier, Microsoft’s Claude-based agentic tool for handling multi-step tasks