Microsoft unveils “Humanist AI” code, putting humans back in control

On September 14, Microsoft introduced a new “Humanist AI” code of conduct that will govern how it builds and deploys artificial intelligence systems across the company. The document, which applies to large language models and other generative AI tools, states that people must retain “meaningful control” over any AI‑driven process and that systems should be explainable enough for users to understand why a particular output was produced.

Under the code, harmful or biased behavior from an AI model is treated as a failure of the overall system, not as a “choice” made by the AI. Responsibility is placed squarely on Microsoft and customer organizations operating the tools. The company also plans to release practical checklists aimed at IT departments, public agencies and smaller firms so they can assess risks before rolling out AI at scale.

The move comes as U.S. policymakers argue over how tightly to regulate AI, with the Trump administration signaling skepticism toward strong federal rules. By publishing its own guardrails, Microsoft is positioning itself as a first mover on responsible AI, regardless of how quickly legislation advances. For enterprises, the code is a signal that governance and accountability expectations around AI are about to rise.