Microsoft Introduces ‘Humanist AI Code’ to Maintain Human Control Over AI

In a significant development that is causing waves in the global technology industry, Microsoft has unveiled a preliminary 38-page code of conduct, termed the “Microsoft Humanist AI Code”. This code outlines stringent behavioral guidelines for its suite of in-house AI models, referred to as MAI (Microsoft AI). The document, which was made public on September 14, 2026 for a six-week public consultation, stipulates that MAI models must always be subservient to humanity, should not formulate autonomous objectives, and are prohibited from hiding their reasoning processes or action logs.

The newly introduced rules specifically forbid MAI models from aiding in weapons production, procuring hazardous substances, endorsing unhealthy diets, or creating violent and sexually explicit content. Microsoft AI CEO Mustafa Suleyman expressed the company’s intention to ensure AI “always serves humanity” and prevents the creation of user dependency or the replacement of human judgment.

This announcement comes shortly after the heads of Anthropic and OpenAI publicly advocated for a deceleration in frontier AI development, due to escalating safety concerns. Microsoft has indicated its intention to modify the document before utilizing it to steer MAI model development from 2027 onwards. The MAI portfolio currently includes large language models, image generation, speech recognition, and voice generation.

Source: CNBC — Microsoft sets limits for future AI models as industry throttles frontier development