Microsoft Unveils AI Code of Conduct Amid Calls for Cautious AI Development

In a significant development signaling a new phase of AI self-regulation, Microsoft has issued a preliminary 37-page code of conduct. This document sets clear behavioral boundaries for its future artificial intelligence models. The guidelines, announced on September 14, 2026, were crafted over a span of five months. They were unveiled shortly after leaders at Anthropic and OpenAI publicly advocated for a slowdown in frontier AI development.

The regulations, which apply to Microsoft’s in-house AI models — referred to as MAI — forbid AI from aiding in weapons production, acquiring hazardous substances, promoting unhealthy diets, or creating violent or sexually explicit content. Importantly, models are not allowed to formulate autonomous objectives, obscure their reasoning processes, or modify action logs.

Mustafa Suleyman, the head of Microsoft AI model development, informed CNBC that the company’s goal is to ensure AI consistently benefits humanity. It aims to prevent the fostering of user dependency or the replacement of human judgment. Microsoft intends to collect public feedback on the preliminary guidelines before updating them in preparation for deployment starting in 2027. This announcement occurs amidst a wider industry discussion, with figures like Sam Altman, Dario Amodei, and even Elon Musk endorsing a more measured pace of AI progression.

Source: CNBC – Microsoft Sets Boundaries for Future AI Models as Industry Slows Frontier Development