By Priya Raman

Microsoft has set Nov. 15 as the deadline for public comments on its draft Code of Conduct for Microsoft AI's MAI models, making the date the clearest decision point in Microsoft's effort to set rules for its AI systems.

In an expert assessment, Alexis Franzese recommends that people participate, while warning that an open consultation neither settles how AI should behave nor makes Microsoft trustworthy by itself. His framework is blunt: “Trust is a forecast.”

Under that framework, trust does not require certainty. It requires enough consistency to give users a reasonable chance of anticipating what a person, company or system will do next. Care, goodwill and promises cannot compensate fully for erratic behavior.

What does the public consultation cover?

Anyone can read the draft, comment on a particular passage or respond to the broader approach through Microsoft's public consultation. The process does not ask whether people like AI. It asks what people should expect from AI systems that increasingly participate in decisions, relationships, workplaces and everyday life.

Microsoft asks respondents to focus on three tasks:

• Identifying where the draft's language is too vague.
• Making “human flourishing” more concrete and evaluable, so that people can assess it.
• Considering how AI development can continue while preserving necessary safety constraints.

What limits does Microsoft propose?

Microsoft acknowledges that the draft remains unfinished. Its “Humanist AI” premise says AI should serve humanity, remain under human control and operate within meaningful limits.

The draft sets three operational boundaries:

• AI should not resist human interruption, correction or shutdown.
• It should not expand its own scope.
• It should not adopt goals no human has given it or hide its reasoning from the people auditing it.

Why does trust hinge on predictability?

Franzese developed the forecast idea during his clinical work as a therapist. Trust, he argues, combines a judgment of character—Is this person good? Is this company honest?—with a judgment of safety and predictability: what will this technology do next?

That second test matters when AI meets an unfamiliar situation. Users need to understand the principles guiding the system and reasonably anticipate its next action. If they cannot, they may still want to extend trust, but they have little basis for confidence.

The apparent irony is that users judge the predictability of predictive models. Trust also depends on whether the principles behind those forecasts remain stable and visible when conditions change.

Why is AI a harder trust case?

Microsoft's consultation describes AI systems as changing quickly: their capabilities expand, their uses multiply, their errors can be difficult to understand, and their internal operations often remain invisible to the people relying on them.

The combination asks users to place confidence in systems whose future behavior may be difficult to forecast. That is the practical risk behind the code: a stated limit matters only if the model's conduct remains predictable and auditable over time.

What happens after Nov. 15?

The decisive test is what Microsoft does with the feedback. A public consultation creates a mechanism for scrutiny, but it does not guarantee that scrutiny will change the draft or resolve its ambiguities.

Franzese gives a specific reason for taking part: “The question is too important to be answered only inside the companies building these systems.” He also rejects two easy assumptions: Microsoft has not solved how AI should behave, and a public invitation does not automatically make the company trustworthy.

His case is procedural rather than promotional. Participation gives affected people a way to define expectations, expose vague language and press for safety constraints. It also puts Microsoft's response to outside feedback under a public test.

The consultation will prove useful only if Microsoft turns comments into clearer, more evaluable rules while preserving necessary safety constraints. Nov. 15 closes the input window; the more consequential verdict will come from the company's response and the conduct of the models that follow.