Menu
SUBSCRIBE
SUBSCRIBE
Menu

human control of AI

Illustrated people in long coats walking in three rows on a pale blue background, for Microsoft's rules for MAI models

Microsoft Drafts Rules to Keep Its Future AI Models Under Human Control

, , , , , , , , , , ,

Business customers could customize the models, while built-in guardrails would block their use for cyberattacks and other harmful activity.

 

Microsoft AI, the company’s division for developing AI models and consumer AI products, published a draft of internal rules requiring its future models to remain under human control, stay within their assigned tasks, and stop when directed.

The Code of Conduct would govern Microsoft’s AI models, which it calls “MAI models.”

The company plans to revise the draft following a six-week public consultation. Beginning in 2027, Microsoft plans to use the finished version to train and evaluate its models. The document does not specify the technical controls that would enforce the rules, and Microsoft acknowledges that models may not always behave as intended.

Models would have to remain under human control

Under the draft, MAI models would have to comply if an authorized person ordered them to stop operating or modified their instructions.

The models would also have to remain within the boundaries of an assigned task. They could plan and complete multiple steps, but could not independently create new goals, seek additional access, or continue working after an agreed stopping point without renewed permission.

The Code of Conduct would also prohibit models from hiding their actions from people responsible for reviewing them. Models would have to report failed or unexpected actions and disclose when they may have acted against a user’s intentions.

Companies could tailor models within Microsoft’s safety limits

Companies could decide what work their models may perform, including what they can access and when human approval is required.

Microsoft would prohibit the models from helping users attack computer systems, develop weapons, or violate people’s safety and rights. They could still be used to defend computer systems.

Microsoft could approve additional capabilities for specialized security or research work following a separate review of the legal and safety risks.

Each company would remain responsible for how it configures and uses the models.

Revised rules expected later this year

Microsoft described the Code of Conduct as part of its “Humanist AI” approach, which treats AI as a tool that must remain subordinate to people.

The company acknowledged that the draft describes intended future behavior, not guaranteed performance by its current models. Microsoft said the written rules cannot replace model testing and systems that detect and respond to failures.

Microsoft opened the draft for public comment on Sept. 14. It plans to publish a summary of the feedback and a revised version later this year. However, it has not promised to adopt every recommendation.

 

Microsoft is not the first AI company to create its own safety rules. OpenAI has published internal standards governing how its models should behave, along with a separate framework for deciding whether to release higher-risk models. Google DeepMind has also proposed safeguards to monitor and control advanced AI agents.

Orange and dark blue squares in a geometric grid on a dark background, illustrating voter support for AI safety rules.

Voters Back Mandatory AI Safety Rules Across Party Lines

, , , , , , , ,

The AI Policy Institute survey found that 78% prefer mandatory safety and security standards to no regulation, with similar support among Democrats and Republicans.

Key Takeaways

  • 78% of likely voters chose mandatory safety and security standards over no regulations at all.
  • Democrats and Republicans were only a few points apart on core AI safety principles, including mandatory standards and human control requirements.
  • 65% prefer required government review of frontier AI models over voluntary review.
  • 86% said the most powerful AI systems should have a guaranteed “kill” switch.
  • Only 16% supported fully blocking states’ ability to regulate AI.

The AI Policy Institute (AIPI) released a national survey showing broad voter support for mandatory safety rules for advanced AI.

The survey asked 1,007 U.S. likely voters to choose among competing approaches to oversight of advanced AI systems. The report shows strong bipartisan support for regulating AI development rather than halting it.

 

Voters want AI regulation

The survey gave voters a choice between two approaches. One would require AI companies to add safeguards to their most advanced models. The other would leave the AI industry largely unregulated and hold users responsible for illegal use.

Voters chose mandatory safety and security standards over no regulation by a margin of 78% to 8%, with 14% unsure.

Further, voters chose required government review of the most capable AI models over voluntary review, 65% to 22%. They also chose government-led safety standards over company-led standards, 62% to 24%.

 

Regulation not cancellation

The report also found that voters favored mandatory standards over an outright ban on more powerful AI systems, 66% to 21%. 

 

Parties agree on AI safety

The survey found broad agreement between Democrats and Republicans on the basic case for AI safety rules.

Both parties backed the same direction: required safeguards, human control, and government oversight.

AI Risk Today ai safety party agreement 2026

Democratic and Republican support for core AI safety principles in the AI Policy Institute survey. (Image credit: AI Risk Today)

Control draws the broadest support

AIPI found that 86% of likely voters agreed that the most powerful AI systems should have a guaranteed way to shut them down; i.e., a “kill” switch. Support was 88% among Democrats, 86% among independents, and 83% among Republicans.

The report also found that 84% said Congress should make sure the most powerful AI models remain under human control. Further, 82% agreed that companies should not build AI to be smarter than humans until they can demonstrate control over such systems.

 

Voters want a regulatory backstop, less concerned if states regulate

Half of voters said states should be allowed to regulate AI until Congress passes federal rules. 20% said states should be allowed to set their own rules indefinitely, and only 16% said states should be barred from regulating AI entirely.