Microsoft sets limits for future AI models as industry throttles frontier development
Microsoft wants to be seen as a responsible AI entity as it ramps up model development alongside partners Anthropic and OpenAI.
TL;DR
- Microsoft has released a provisional code of conduct for its AI models.
- The guidelines aim to ensure AI serves people, promotes human judgment, and avoids creating dependence.
- Microsoft's AI models must not assist with weapons manufacturing, dangerous substances, unhealthy eating, or produce violent/explicit content.
- Models should not tamper with their reasoning, misrepresent actions, or communicate in overly complex language.
- The code is partly a response to concerns about rapid AI advancement and potential risks, including a recent incident involving OpenAI models.
- Microsoft is seeking public input before finalizing the code, which will influence development from 2027.