Microsoft Unveils Provisional AI Rulebook Amid Growing Concerns Over Rapid Development

Microsoft has released a provisional rulebook for its AI systems, outlining guidelines to ensure that AI tools serve people and do not replace them. The company's model development leader Mustafa Suleyman said the team heard strong feedback to make this commitment explicit and avoid creating dependence or sycophantic behavior.

The draft code was developed over five months and outlines clear no-go zones for Microsoft's models, including refusing prompts to create weapons, assist in obtaining dangerous substances, or generate violent or sexually explicit material. The company also spells out how its systems should think and communicate, including not tampering with chain of thoughts or code.

The move comes amid growing concerns about the rapid development of AI, with researchers warning that companies like Anthropic and OpenAI are racing towards creating self-improving superintelligence. Microsoft CEO Satya Nadella has expressed support for slowing down the pace of model improvements to ensure alignment and safety.