Microsoft has declared that AI "is not conscious and should not be designed to imitate consciousness", marking out their working position on an increasingly live question among frontier AI lab researchers.
The company sketched the stance in an 11,000-word draft ‘Code of Conduct’ its researchers should follow when training and operating in-house MAI models and, by extension, the "values and behaviour" which the models themselves ostensibly display.
The treatise, which is not yet being used for training and is subject to a six-week public consultation, maps out a philosophy of "Humanist AI" and lands amid calls from leaders at Anthropic, Open AI, xAI and Google DeepMind to ‘pace the frontier’ of AI development in order to mitigate risks of agentic loss of control and other would-be cataclysmic scenarios.
Microsoft themselves have not explicitly called for a slowdown, but Mustafa Suleyman, CEO of Microsoft AI, said that "fears about possible loss of control are real" in a post accompanying the Code of Conduct’s release on Monday.