John Langford says he has left Microsoft and that AI Frontiers, the research group he helped form, is over. In a November 6, 2026 post, he describes a September move into Microsoft AI followed by a dispute over research priorities. For developers using the group's software, his account leaves a separate question: what support can they expect?
The post is a participant's account, rather than a Microsoft announcement. Langford's criticism of Microsoft AI's leadership is his assessment. It does not establish that Microsoft has stopped AI research across the company.
A disagreement about how to improve models
Langford says AI Frontiers formed within Microsoft Research in 2023 to organize work that individual researchers could struggle to deliver alone. He identifies AutoGen, a framework for coordinating AI agents, among its projects. He describes Microsoft AI's focus as scaling models and refining data, and says that conflicted with his interest in developing new capabilities. His departure post also says he could not get his research into use there.
His account connects the disagreement over research priorities to the difficulty of getting that work used. It supplies no company-wide research budget or staffing breakdown with which to measure a broader shift.
AutoGen users have a separate maintenance decision
The supplied AutoGen README says the project is archived and no longer actively maintained by Microsoft. Microsoft says it is not providing bug fixes, security fixes, new features, or support for this project. The README also advises against production use. It directs new users toward Microsoft Agent Framework and encourages existing users to migrate. These statements do not date the archive decision or tie it to Langford's departure.
Lower sections describe AutoGen as community-managed and say contributions in maintenance mode are limited to bug fixes, security patches, and documentation improvements. Those statements can coexist with Microsoft's withdrawal of its own maintenance and support. The supplied documentation does not establish how active community maintenance is or when users should expect a response.
For an existing dependency, the useful next step is to assess whether community maintenance meets the application's support needs and review the linked migration guidance. Microsoft's recommendation alone does not establish that a migration will work for a particular application. The supplied evidence contains no framework or migration test results.