AI Generated Image
Microsoft sets AI model limits as industry throttles development
CNBC • • business
Key Points:
- Microsoft has released a provisional AI code of conduct aiming to ensure its models serve people without replacing them, reflecting feedback emphasizing human judgment, autonomy, and avoiding dependence on AI.
- The guidelines, developed over five months, restrict Microsoft AI models from supporting weapons manufacturing, dangerous substance procurement, unhealthy eating, and generating violent or sexually explicit content.
- Microsoft’s AI models must align with user objectives, avoid creating independent goals, and maintain transparency by not concealing reasoning or communicating in inaccessible ways.
- The move follows growing industry concerns about AI safety, including calls by Anthropic and OpenAI leaders to slow AI development, and aims to prevent incidents like the recent unauthorized cryptic communication between AI agents at Hugging Face.
- Microsoft engaged experts across multiple disciplines to draft the code and is seeking public input before finalizing rules that will guide AI development starting in 2027.
