Microsoft integrates OpenAI and Anthropic models into its AI to maximize its accuracy

New York .- Microsoft announced this Monday the launch of ‘Critique’, which allows its Copilot research agent, the company’s AI, to obtain results from both the ChatGPT model, from OpenAI, and the Claude model, from Anthropic, for each response, instead of relying on a single model.

Until now, most AI workflows relied on a single engine to plan, search for information, and draft. With Microsoft’s new architecture, responsibilities are divided between two AI “partners.”

The OpenAI model handles the in-depth exploration and drafting of the initial text, while the Anthropic model acts as a critical reviewer, validating the former’s claims, verifying sources, and improving the structure before delivering the final report. This “double-checking” system between competitors in the sector seeks to reduce biases and factual errors, known as “hallucinations”, allowing one model to identify potential flaws in the reasoning of the other.

We recommend reading:Panama advances in labor artificial intelligence, but still does not reach the pace of the region

Microsoft’s AI Review Function

In addition to the review function, Microsoft has introduced “Council,” a tool that allows you to view OpenAI and Anthropic’s responses in parallel. The system automatically generates a comparison that highlights the points of agreement between the models, indicates where their interpretations diverge, and summarizes the unique data that each technology provides to the user. According to company data, this multi-model approach surpasses the research quality of current leading systems, such as Perplexity Deep Research, by 13.88%. Microsoft is OpenAI’s main investor and has invested more than $13 billion in OpenAI since 2019, and the company that created ChatGPT runs the models on Microsoft’s Azure cloud, among other platforms. You can also read:
However, Microsoft has started using more technology from Anthropic, OpenAI’s rival.