Microsoft Publishes Draft Humanist AI Code of Conduct
Microsoft has published its draft of the Humanist AI Code of Conduct, which aims to guide future training and model behavior for its MAI systems. The code emphasizes human supervision and enforceable restrictions as key factors in evaluating AI systems.
The draft's most important objective is to ensure that humans remain in control of MAI systems, which should perform more like tools than people. It includes Absolute Constraints that cannot be violated, such as using weapons or causing mass harm, engaging in cyber offensive operations, loss of human control, large-scale malicious manipulations, and child safety matters.
The code also establishes a three-level Chain of Command: the Code of Conduct takes top priority, followed by operator's policies, and user's preferences. This chain ensures that Absolute Constraints and Human Control Requested prevail and cannot be neglected.
MICROSOFT said that the draft is not used in training models today and will publish a revised version later this year to facilitate MAI development by 2027 and beyond.