Microsoft has released a draft Code of Conduct for its MAI models that will serve as the primary governing document for development from 2027 forward, opening a six-week public consultation to shape the final version.
The framework, developed through consultations with AI experts, legal advisors, ethicists, philosophers, linguists and business leaders, codifies Microsoft's approach to what it calls Humanist AI—a system design that retains meaningful human control over increasingly capable systems.
The code defines helpfulness as the degree to which a response assists users in accomplishing their goals, weighting user-perceived value alongside factors like localization quality and creativity rather than objective correctness alone. This metric matters for product design: it signals Microsoft is optimizing for adoption and end-user satisfaction in enterprise and consumer applications, not just technical benchmarks.
Microsoft plans to use this framework to evaluate MAI model performance and to inform technical controls, operational monitoring systems and internal training practices. The revised document, due by year-end, will directly guide model development beginning in 2027.
The move reflects competitive positioning within frontier AI. Anthropic and OpenAI have advocated publicly for slower AI development, staking claims to trustworthiness and safety. Microsoft's structured governance framework serves a similar strategic function—demonstrating to regulators, enterprise customers and policymakers that the company is taking human oversight seriously.
The company has invited input from users, operators, researchers, governments and the general public, signaling an intent to expand stakeholder engagement as new code iterations develop.
