Microsoft proposes limits on its AI with code of conduct amid safety debate

Microsoft has unveiled a provisional “code of conduct” for training new artificial intelligence models, aiming to ensure AI remains subordinate and beneficial to humanity. CEO of Microsoft AI, Mustafa Suleyman, emphasized the urgent need for safety guidelines as concerns grow regarding the potential loss of control over autonomous systems. The proposed rules prohibit AI from assisting in weapons development, generating violent or explicit content, or imitating human consciousness. This move follows rising industry anxiety, with leaders from Anthropic and OpenAI also calling for increased safety measures and third-party oversight. While some experts welcome the initiative, others remain skeptical, warning that such standards could entrench the dominance of major labs and prove ineffective without global cooperation. The guidelines are currently open for public consultation as the tech sector grapples with the rapid, and potentially risky, evolution of frontier AI models.