Microsoft AI CEO releases Humanist Code of Conduct for AI models; calls it 'urgent'

Microsoft unveiled a draft code of conduct for its internal artificial intelligence (AI) models on September 14, 2023. The guidelines aim to ensure that increasingly capable AI systems remain subordinate to human control, according to Microsoft AI CEO Mustafa Suleyman. Suleyman stated, "AI must be subordinate and always in service of the people."

The draft policy tackles concerns within the tech sector about autonomous models or AI agents that may cross dangerous thresholds while performing assigned tasks. It comes shortly after Anthropic CEO Dario Amodei released an open letter urging the industry to slow down AI development.

Under the proposed guidelines, Microsoft's automated systems must never contradict human corrections or shutdown commands, should use easily understandable language, and must treat any breach of these behavioral rules as a complete system failure. Suleyman emphasized, "People matter more than AI. AI must be subordinate and always in service of people."

The draft consists of 10 points, urging feedback from readers over the next six weeks. Suleyman outlined the key points, including the importance of placing people above AI, rejecting the notion of AI model welfare, and ensuring that models never meaningfully violate the code. If a model finishes its job or breaches the code, it is considered a failure.