Microsoft Ties AI Models to Human Authority in New Code of Conduct

Microsoft's Humanist AI code of conduct sets human control as a hard limit for its artificial intelligence (AI) models. The company states that its models must remain subordinate to people at all times and cannot set their own goals, resist human input, or be designed as autonomous agents.

The 37-page document, which will inform model development starting in 2027, marks a significant concession by Microsoft. CEO Satya Nadella voiced support for slowing AI development, following Anthropic researcher Jacob Coxon's resignation and public warning that AI companies are 'gambling with human lives.'

Microsoft AI chief executive Mustafa Suleyman told Axios he does not accept the view that superintelligent systems will inevitably slip beyond human authority. He argued that a slower or less capable model is preferable to one that cannot be controlled.

The code also rejects any pursuit of legal personhood for AI systems and dismisses the idea that models could be entitled to rights or welfare. Suleyman called the idea of designing AI to imitate consciousness 'very, very dangerous.'