# Microsoft Unveils "Humanist AI" Code of Conduct, Bars Models From Resisting Shutdown

*2026-09-14 22:01 UTC · AI · 1 min read*

> Microsoft published a draft "Humanist AI" code of conduct requiring its models to never resist shutdown or human control, opening a six-week public feedback period before training its next generation.

Microsoft published a draft code of conduct for its in-house AI systems, the product of five to six months of work. AI chief Mustafa Suleyman called it a "constitution" the company will fold into training its next generation of models.

The document sets four mandatory constraints: models must never resist human interruption, override, redirection or shutdown; they cannot assist with weapons development or cyberattacks; they must reject any claim to simulated consciousness, feelings or rights; and any task that would violate the code must fail outright rather than be completed some other way.

Microsoft frames the approach as "Humanist AI" — AI kept subordinate to human users — explicitly rejecting the industry race toward an all-purpose superintelligence. "We reject the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights," the company said.

Suleyman opened a six-week public comment period Monday, saying "Now's the time for coordination, and coordination means disclosing how capable your models are to responsible third parties." Microsoft plans to use the finished code to train its next generation of models once the window closes.

## Sources

- [fortune.com](https://fortune.com/2026/09/14/microsoft-suleyman-ai-safety-code-of-conduct/)
- [betanews.com](https://betanews.com/article/microsoft-ai-code-shutdown-resistance/)

---
TechTools News · https://techtools.cz/tools/news/article/2373-microsoft-unveils-humanist-ai-code-of-conduct-bars-models-from-resisting
Česká verze: https://techtools.cz/tools/news/clanek/2446-microsoft-predstavil-kodex-chovani-humanist-ai-zakazuje-modelum-vzdorovat