Microsoft Proposes Framework to Keep AI Models Under Human Control

Microsoft has proposed a framework to ensure its AI models remain under human control. The draft code of conduct, open for public consultation until November 2023, outlines rules for transparency and limits on AI capabilities.

The company's 'Humanist AI' concept emphasizes the subordination of AI systems to humans rather than operating independently. Microsoft's AI chief Mustafa Suleyman describes the document as a governing framework for future models developed by the company's AI division.

The proposed code requires Microsoft AI models not to resist human correction, interruption, or shutdown and sets expectations around transparency. It establishes a hierarchy governing how the models should respond to instructions from Microsoft, organizations deploying them, and individual users.