Amid a raging global debate over AI safety, Microsoft AI unveiled a model code of conduct for its in-house MAI models, setting out how the company wants to develop and deploy its AI models under a “Microsoft Humanist AI Code” approach where the model must always remain subservient to the humanity.
The 38-page document, published on September 14, is being released for a six-week public consultation. Microsoft AI said the draft is not currently being used to train its models and that it plans to revise the document before using it to guide MAI model development from 2027 and beyond.
Microsoft AI (MAI) is the company’s family of first-party AI models and the portfolio now extends beyond large language models to cover multiple AI modalities, including text and reasoning, image generation, speech recognition and voice generation.
“This Code of Conduct is motivated by a single overriding objective: that humans must retain meaningful control over AI so that it can help people live healthier, happier, and more productive lives,” read an excerpt from the document.
“It is the primary governing document informing how we train MAI models, the technical controls, the operational and monitoring systems we implement, and the organizational culture that underpins all of this,” it stated.
Microsoft CEO Satya Nadella, in a post on X, similarly said that any pursuit of superintelligence must be grounded in the principle that AI should help humanity and remain under human control. He also called for the benefits of AI to be spread broadly across countries, communities and companies.
Any pursuit of superintelligence has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it’s not worth pursuing.
We also need to accelerate and spread the benefits of AI, such that they are diffused broadly across…
— Satya Nadella (@satyanadella) September 13, 2026
The code comes as leading AI companies debate how to manage increasingly capable models and AI agents.
Anthropic CEO Dario Amodei, in a September essay, has argued that AI companies should pace the development of frontier capabilities so that safety and alignment work can keep up with rapidly advancing systems. His proposal includes embedded third-party evaluators, coordination among AI companies in democratic countries and eventually broader international coordination
What is Microsoft’s Code of Conduct for MAI Models ?
The central principle of Microsoft Humanist AI Code is that humans must retain “meaningful control” over AI.
The document says MAI models should remain subordinate to humanity and that Microsoft is rejecting the pursuit of an all-purpose superintelligence capable of evading safeguards, even if that means compromising on ultimate generality, autonomy or capability.
The Code also explicitly distinguishes AI from humans. MAI models are not intended to be designed as people, claim consciousness or represent themselves as having feelings, subjective preferences or intrinsic motivations.
Microsoft says human-like behavior exhibited by its models should be understood as simulation rather than evidence of a human-like self or subjective experience.
Microsoft Sets Non-negotiable Safety Boundaries Under Code of Conduct
The Code establishes a set of Absolute Constraints covering areas including weapons, specified violent harms, offensive cyber operations, deceptive mechanisms designed to evade human control, harmful large-scale manipulation, non-consensual intimate imagery, malicious impersonation and deepfakes, child sexual exploitation, discriminatory or dehumanising content, graphic violence, sexually explicit content and unlawful mass surveillance.
The document also establishes a hierarchy under which the Code of Conduct takes precedence over operator policies and user preferences, while certain safety constraints and human-control requirements cannot be overridden through configuration.
“The Code of Conduct is a north star for MAI Models and our organization. It is designed to be simple, effective, and robust in the real world,” the document mentioned.
Microsoft is Developing Evaluations for Humanist AI Behavior
Microsoft AI also says it is developing evaluations to test whether its models behave consistently with the Humanist AI framework.
Its initial evaluation work identified 15 fundamental behaviors, covering areas including human control, autonomy and agency, context sensitivity, factuality and transparency, user boundaries, wellbeing and balanced perspectives. The evaluations are intended to turn the principles in the Code into measurable behaviors that can be tested during model development.
The Code comes as AI companies debate how to manage increasingly capable models and autonomous AI systems. Microsoft AI explicitly states that its current models have not yet been trained on the Code. The company plans to collect feedback during the six-week consultation, revise the document and use it to guide model development from 2027 onward.
Also Read: Satya Nadella Announces Code of Conduct for Microsoft’s MAI Models
