Microsoft has struck a licensing agreement with Harvard Medical School’s publishing arm to grant its Copilot AI assistant access to consumer health content covering diseases and wellness topics: a move designed to improve the credibility of AI health responses. In return, Microsoft will pay the Ivy League University a licensing fee.
The collaboration comes as Microsoft seeks to reduce its reliance on OpenAI’s models, which currently power much of Copilot. The company is already diversifying: it added already support for Anthropic’s Claude models to 365 Copilot. In licensing Harvard’s content, Microsoft aims to make Copilot’s health-related answers closer to what users might hear from a medical professional.
The Harvard deal addresses a key weakness in consumer AI: trust. A Stanford study, for example, found that ChatGPT gave inappropriate medical advice in about 20% of cases. By anchoring responses in a respected medical publisher, Microsoft hopes to build a “vertical moat” in healthcare, where content credibility becomes a differentiator.
On Microsoft’s side, the move also fits broader technical strategy. The company is building its own AI stacks, integrating new models, and slowly shifting work away from the OpenAI backbone. Meanwhile, Microsoft has published research suggesting its diagnostic AI outperforms human physicians in complex cases, claiming accuracy rates four times higher in experimental settings.
Still, risks remain. AI models can hallucinate or misinterpret medical nuance, and licensing Harvard’s content is not the same as clinical validation. Microsoft, Harvard, and regulators will need to monitor how Copilot surfaces that information in practice and when lines between information and medical advice become legally or ethically relevant.
