Microsoft Publishes Draft of Humanist AI Code of Conduct

Microsoft has taken a significant step in establishing guidelines for its AI models. The company has published a draft of its Humanist AI Code of Conduct, which outlines how it develops and intends to deploy its AI models. This training manual is intended for Microsoft's model-training teams, users, organizations operating its models, researchers, governments, and the public.

The draft document is open for public consultation for six weeks. After reviewing the feedback, Microsoft will revise the document and publish an updated version later this year. The revised Code of Conduct is expected to guide model development from 2027 onward.

Microsoft AI's current models have not been trained on the Code, which remains under development. The Humanist AI Code of Conduct builds on Microsoft's concept of 'Humanist Superintelligence,' an approach designed around human needs and control. It aims to provide standards for developing, training, and evaluating Microsoft AI models.

The framework establishes restrictions intended to prevent models from enabling serious harm. For instance, it prohibits assistance involving chemical, biological, radiological, nuclear, and explosive weapons. Models must also remain under human direction and operate within defined limits.