Microsoft Lets One AI Model Critique Another in Copilot Upgrade

Microsoft has introduced a new feature in its Copilot AI assistant that allows multiple AI models to work together and review each other’s responses, in a move designed to improve accuracy, reduce AI errors, and increase productivity for users.

The new feature, called “Critique,” is part of Copilot’s Researcher agent. With this feature, Copilot will no longer rely on a single AI model to generate responses. Instead, it will use two different AI models from different companies in the same workflow. In this system, OpenAI’s GPT model generates the response, and Anthropic’s Claude model reviews the response for accuracy, quality, and possible errors before the final answer is shown to the user.

Microsoft says this multi-model approach will help reduce AI hallucinations, which occur when AI systems generate incorrect or misleading information. By allowing one AI model to review another model’s work, Microsoft aims to produce more reliable and higher-quality results.

The company also plans to make this process bi-directional in the future, meaning GPT will also review Claude’s responses, creating a system where AI models check each other’s work. This collaborative approach is designed to combine the strengths of different AI systems.

Microsoft executive Nicole Herskowitz, Corporate Vice President of Microsoft 365 and Copilot, said that having multiple AI models working together gives customers better results because each model has different strengths. By combining them, users can benefit from improved accuracy, faster research, and better overall output quality.

In addition to the Critique feature, Microsoft is also launching a new feature called “Council.” This feature will allow users to compare responses from multiple AI models side-by-side, helping users evaluate different answers and choose the best one.

These updates are part of Microsoft’s broader effort to improve its AI assistant and increase adoption of Copilot. The company is also expanding access to its new Copilot Cowork tool, an autonomous AI agent designed to help users complete complex tasks and workflows. This tool is currently being made available to users in Microsoft’s Frontier program, which gives early access to new AI features.

Microsoft is rapidly improving Copilot as competition in the AI industry increases. Major competitors include Google’s Gemini and autonomous AI tools such as Claude Cowork. By introducing multi-model collaboration and AI review systems, Microsoft is trying to make Copilot more reliable, accurate, and useful for business and professional users.

This development represents a new trend in artificial intelligence where AI systems do not work alone but instead collaborate, review, and improve each other’s outputs, leading to more trustworthy and efficient AI tools.