This review examines the introductory structure of Part 2, “Safety,” in Microsoft's Humanist AI Code of Conduct. The section establishes the safety constraints governing MAI Models and introduces several topics intended to guide readers through the document.
The editorial observations below concern the relationship between this introduction and the sections that follow, particularly the consistency of its hierarchy, the completeness of its outline, and the distinction between subsections of Part 2 and references to other parts of the Code.
On “Safety”: Introductory Structure and Hierarchy
Introductory structure and section headings
The opening paragraph of Part 2 suggests that the introduction will provide an outline of the Safety section.
However, the five bold topics that follow do not correspond directly to the five section headings that organize Part 2.
This may create uncertainty about whether the introduction is intended as an outline of the section, a summary of selected safety principles, or an overview of the document's broader structure.
Incomplete coverage of the Safety section
The five introductory topics refer to three of the five principal sections within Part 2:
- Chain of Command
- Absolute Constraints
- Operator Configurability
Two sections are not represented:
- Operationalizing Safety
- Human Control
Although an introduction does not necessarily need to enumerate every subsection, the current presentation may lead readers to expect a complete outline.
Consider whether these omissions are intentional and whether the introduction should establish more clearly how the five sections relate to one another.
Mixing hierarchical levels
The final two bold topics in the introduction are Guidelines and Defaults.
Unlike the preceding topics, these refer to Part 3, “Operational Guidelines,” and Part 4, “Operational Defaults,” rather than subsections of Part 2.
Presenting them alongside Part 2's own subjects may blur the distinction between different levels of the document's hierarchy.
Consider separating the overview of Part 2 from references to other parts of the Code, or clarifying why these topics are presented together.
References
- Microsoft AI. (2026, September 14). Humanist AI Code of Conduct. Part 2, “Safety,” pp. 9–15.
