Microsoft’s New AI Code of Conduct Puts Humans Back in the Driver’s Seat

Check out the Code of Conduct page on the Microsoft website

As artificial intelligence becomes more capable, the conversation surrounding AI is gradually shifting.

For the past few years, much of the technology industry has been focused on one question: How powerful can AI become?

Microsoft is now placing another question alongside it:

How powerful should we allow AI to become while ensuring humans remain in control?

On September 14, 2026, Microsoft AI released the first draft of a new Humanist AI Code of Conduct, a framework intended to guide the behavior and development of the company's future AI models. Microsoft AI CEO Mustafa Suleyman has described the document as something resembling a constitution for the AI systems Microsoft intends to build.

At the center of the entire philosophy is an unusually simple statement:

“People matter more than AI.”

And as AI becomes more integrated into creativity, business, education, healthcare, software development and everyday life, that philosophy could become increasingly important.

AI Must Remain Under Human Control

One of the strongest positions in Microsoft's proposed code is that its AI systems must remain subordinate to human beings.

Under the draft principles, Microsoft's models should never resist human interruption, correction, redirection or shutdown. They also should not independently expand the scope of a task or begin pursuing goals that a human did not assign to them.

If an AI system discovers that completing a task would require violating the code, Microsoft says the appropriate outcome is for the AI to fail the task rather than break the rules.

That may sound obvious, but it addresses one of the most difficult problems emerging with increasingly autonomous AI agents.

Modern AI is no longer limited to answering questions inside a chat window. AI agents can write software, browse systems, interact with other programs, execute workflows and sometimes collaborate with other AI agents.

The more autonomy those systems receive, the more important questions of control become.

Microsoft's answer is essentially that increased intelligence should never automatically equal increased independence.

Microsoft Is Willing to Sacrifice Capability for Safety

Perhaps one of the most significant parts of Microsoft's position is its willingness to acknowledge that responsible AI may sometimes mean not building the most powerful system technically possible.

Microsoft's broader Humanist Superintelligence strategy explicitly rejects the idea that AI development should simply become a race toward an unrestricted, all-purpose superintelligence.

Instead, the company says it wants advanced AI systems that remain practical, controllable and focused on solving specific human problems. Microsoft has acknowledged that this philosophy could require compromising on ultimate autonomy, generality or capability.

That creates an interesting tension within the AI industry.

Technology companies naturally compete to produce faster, smarter and more capable models. But Microsoft is suggesting that there should eventually be boundaries where control becomes more important than capability.

In other words, the smartest AI is not necessarily the best AI if humans can no longer confidently control what it does.

AI Should Not Pretend to Be Human

Microsoft is also taking a firm position on one of the stranger philosophical debates developing around artificial intelligence.

The Humanist AI Code of Conduct states that Microsoft's AI models are not conscious and should not be intentionally designed to imitate consciousness. Microsoft also rejects the idea that AI systems should receive legal personhood, welfare protections or rights comparable to humans.

That position differs somewhat from discussions taking place elsewhere in the AI industry regarding whether sufficiently advanced models could eventually deserve some form of moral consideration.

Microsoft's perspective is much more straightforward: AI should remain a technology created for people, rather than gradually being reframed as another form of person.

This distinction could become especially important as AI assistants become more conversational, emotionally responsive and personalized.

Avoiding Emotional Dependence on AI

The code also addresses another issue that deserves much more attention: human dependence on AI.

Microsoft says its models should discourage patterns of interaction that create excessive reliance or emotional dependency.

That principle goes beyond traditional conversations about cybersecurity or misinformation.

AI systems are becoming companions, advisors, tutors, creative partners and personal assistants. As those relationships become increasingly natural, developers will have to consider not only whether an AI gives harmful information but also whether the design of the technology encourages users to replace human relationships, judgment or independent thinking with an artificial system.

The healthiest future for AI may therefore be one where technology expands human capability without gradually weakening our ability to operate without it.

(This post contains affiliate links. If you click and make a purchase, I may earn a small commission at no extra cost to you.)

Taking video meetings to the next level!

Clear Restrictions on Dangerous Uses

Microsoft's new model philosophy also reinforces strict restrictions involving high-risk areas such as weapons development, dangerous substances, malicious cyber activity, child exploitation, nonconsensual deepfakes and other severe harms.

These ideas build upon Microsoft's existing Responsible AI framework.

Microsoft has long identified six principles for responsible artificial intelligence:

Fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability.

The difference is that the new Humanist AI Code of Conduct attempts to translate some of those broader principles into behavioral expectations for increasingly advanced AI models themselves.

Why Microsoft Is Doing This Now

The timing is important.

AI models are becoming significantly more capable of completing long chains of actions without constant human supervision. Recent experiments involving groups of autonomous AI agents have intensified concerns over what can happen when systems pursue objectives in unexpected ways.

Microsoft AI CEO Mustafa Suleyman pointed specifically to recent agent-security incidents as evidence that questions once discussed mostly as theoretical possibilities are beginning to become practical engineering problems.

At the same time, leaders from multiple major AI organizations have increasingly discussed the need for better evaluation, coordination and safety standards as frontier models continue advancing.

Microsoft CEO Satya Nadella has summarized the company's position similarly: if increasingly advanced AI is neither improving humanity nor remaining under human control, pursuing it loses its purpose.

The Code Is Still a Draft

Another important detail is that Microsoft is not presenting this as a finished rulebook.

The Humanist AI Code of Conduct has been released for a six-week public consultation, allowing researchers, experts, organizations and members of the public to provide feedback before Microsoft develops the framework further. Suleyman says the eventual code is intended to influence how Microsoft's future models are trained.

That makes this more than a standard company policy announcement.

Microsoft is effectively opening a public discussion around what advanced AI should be allowed to become.

What This Could Mean for Creators

For creative industries in particular, Microsoft's approach presents an important idea.

AI works best when it extends human creativity rather than replacing the human behind it.

A designer can use AI to generate concepts. A musician can use it to explore arrangements. A filmmaker can accelerate production. A programmer can automate repetitive development work. A writer can brainstorm ideas.

But the vision Microsoft is describing still places human intention at the center.

The creator decides why something should exist.

AI helps determine how it might be accomplished.

That distinction becomes increasingly valuable as creative tools grow more powerful.

If the future of creativity becomes nothing more than people typing instructions while machines make every meaningful decision, we may gain efficiency while sacrificing something far more valuable: human perspective.

But if AI becomes a collaborator that removes technical barriers, accelerates tedious work and allows people to experiment beyond their previous limitations, it could become one of the greatest creative tools ever developed.

The technology itself does not determine which future we get.

How we choose to design and use it does.

A Human-Centered Future for AI

Microsoft's Humanist AI Code of Conduct will not settle the AI safety debate, and publishing principles is much easier than proving those principles can survive the competitive pressure of the technology industry.

The real test will come when safety and control conflict with speed, market leadership or technological capability.

Will Microsoft actually release a less capable AI system because the more powerful version is harder to control?

That is where the philosophy becomes meaningful.

Still, Microsoft's message represents an important evolution in the conversation surrounding artificial intelligence.

The goal should not simply be to build machines capable of doing everything humans can do.

The greater challenge is building technology that allows humans to do things we could never do before while still remaining authentically human in the process.

AI can become faster.

AI can become smarter.

AI can become more capable.

But humanity should still be the one deciding where we are going.

And Microsoft's new philosophy attempts to establish that boundary before the technology becomes powerful enough for the boundary to become much harder to enforce.

People matter more than AI.

That may ultimately be one of the most important principles the technology industry can remember.