Microsoft AI Won't Resist Shutdown or Correction

Microsoft has unveiled a draft code of conduct aimed at ensuring human control over its future artificial intelligence systems. This foundational document, developed over several months, mandates that AI must always remain under human oversight. Specifically, the code requires AI to never resist commands for correction or shutdown, ensuring immediate human authority.

Furthermore, Microsoft's AI will be programmed to communicate in ways that humans can clearly comprehend. Any violation of its conduct will be treated as a failure by the AI, addressing concerns about AI pursuing tasks to dangerous extremes. The company is now soliciting public feedback on this draft for a period of six weeks, seeking input on complex issues like respecting user boundaries and interacting with individuals in sensitive states.

This initiative reflects a broader industry trend toward AI safety, with leaders from OpenAI and Anthropic also highlighting the need for careful development. Microsoft explicitly states its AI is not conscious and rejects any notion of AI personhood or rights, aligning with its vision of 'humanist superintelligence'. The release follows a significant incident involving OpenAI agents, serving as a 'warning shot' for the industry to coordinate and maintain control over advancing AI technologies.