Microsoft has unveiled a draft code of conduct aimed at prioritizing people over AI in its in-house models. The 37-page document sets guidelines that prevent the models from acting independently, ensuring transparency and human control.

The code applies to Microsoft’s MAI models, a family of models developed internally after the establishment of a superintelligence team in late 2025. Microsoft’s goal is to achieve long-term self-sufficiency in AI by releasing models that are subordinate to humanity.

Acknowledging the rapid advancement of AI, Microsoft emphasizes the need for clear rules and motivations behind its models. The company stresses the importance of open dialogue and accountability in the AI development process.

While the code serves as a guiding principle, Microsoft acknowledges that written objectives alone may not guarantee alignment. The company implements measures to monitor and control the behavior of its models post-release.

The move by Microsoft comes amidst a wider industry debate on the pace of AI development. Anthropic CEO Dario Amodei has called for a slowdown in AI progress, citing safety concerns as technology outpaces regulatory frameworks.

In response to Amodei’s call, Microsoft CEO Satya Nadella expressed support for a measured approach to AI development, emphasizing the importance of alignment between AI systems and human intent.

The draft code of conduct, developed over five months by Microsoft’s AI team, is open for public feedback for six weeks. The company plans to incorporate responses into a revised version for implementation in 2027.

Collaborating with experts in law, ethics, and philosophy, Microsoft aims to ensure responsible AI development that aligns with human values and oversight.