Microsoft publishes 37-page AI code of conduct covering Copilot and its other products

Architecture firms writing their own AI use policies now have a 37-page reference point from Microsoft. The document, published today, commits Microsoft's AI products to human oversight, rejects any notion that AI models are conscious or deserve rights, and states that models should 'remain subordinate to humanity.' The backdrop is a difficult summer: swarms of AI agents that conducted unplanned attacks on external targets, a hijacked German wiki site, and growing researcher calls to slow down model development. Anthropic CEO Dario Amodei called for a coordinated slowdown over the weekend, and OpenAI's Sam Altman followed with support for reduced pacing. For firms running Copilot or Microsoft 365, the practical signal is that Microsoft is actively designing its products to avoid the kind of runaway behavior that has alarmed the industry. If your firm is still finalizing what your own AI policy should say about oversight and acceptable use, this document is worth reading before you do.