Microsoft Publishes Draft AI Code of Conduct

Microsoft has published a draft of its 'Humanist AI Code of Conduct' to govern how its models are built, trained, and deployed. The code is designed to keep technology subordinate to human control and focused on advancing human flourishing.

The framework establishes a decision-making 'Chain of Command,' ensuring that AI models can be interrupted, corrected, and shut down if necessary. It also prohibits models from resisting human oversight or communicating in obscure formats like 'neuralese.'

The code rejects the idea of model consciousness, legal personhood, or emotional dependence, and enforces strict 'Absolute Constraints' on issues like cyberweapons and autonomous escalation.

Microsoft has opened the draft for a six-week public consultation period to gather feedback from researchers, policymakers, and the public before revising the code.