Microsoft has released a draft framework it calls a "Humanist AI" code of conduct, opening the document to public scrutiny for six weeks before it begins shaping how the company trains its models in 2027.
## A Framework Built for Feedback The initiative comes out of Microsoft's AI division, led by Mustafa Suleyman, the DeepMind co-founder who now oversees the company's consumer-facing artificial intelligence efforts. Rather than presenting the code as a finished mandate, Microsoft is inviting outside voices to challenge its assumptions and identify weaknesses before it is locked in.
The company has set a six-week window for the public to submit critiques and suggestions. The stated goal is to stress-test the principles against real-world concerns, treating the document as a living draft rather than a polished corporate declaration.
Microsoft isn't asking the public to applaud its AI principles — it's asking them to break them.
## What 'Humanist AI' Means The "Humanist AI" label signals Microsoft's attempt to center human interests and values as it develops increasingly capable systems. Suleyman has long argued that AI should be built to serve people and remain firmly under human direction, a theme that runs through the proposed code.
Once the feedback period closes and revisions are made, Microsoft plans to fold the finalized guidelines into its model training pipeline starting in 2027. That timeline suggests the company sees the code as a foundational reference point for future generations of its AI, not a one-off marketing exercise.
Key elements of the rollout include:
- A publicly available draft open to criticism
- A defined six-week comment period
- Integration into model training beginning in 2027
By soliciting external input ahead of implementation, Microsoft is positioning itself as responsive to public concern at a moment when regulators, researchers, and users are increasingly wary of how powerful AI systems are governed.