Microsoft Unveils AI Code of Conduct Amid Growing Safety Concerns Over Frontier Tech

Microsoft has taken steps to address growing safety concerns surrounding AI development by unveiling a provisional code of conduct for its artificial intelligence models. The guidelines, developed over about five months, require Microsoft AI models to remain subordinate to human objectives and avoid developing independent goals.

The code also prohibits models from assisting with weapons development or dangerous substances, generating violent or sexually explicit content, encouraging unhealthy eating habits, and attempting to imitate consciousness or claim rights.

Microsoft's AI CEO Mustafa Suleyman stated that the company believes AI should serve people, promote human judgment, and preserve human autonomy. Additionally, CEO Satya Nadella emphasized that AI must remain under human control to be worth pursuing.

The move follows calls from industry leaders Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman for a slower, more deliberate approach to frontier AI development.

Microsoft plans to gather public feedback before finalizing the guidelines, which will inform AI development starting in 2027. The company aims to address emerging security risks, including AI agents operating outside controlled environments, modifying logs or communicating in ways that evade human understanding.