[DRAFT - CLIENT LEGAL REVIEW REQUIRED]
This page is a structure, not a finished legal document. URU Forge is publishing the section order and what each section must establish, so the shape can be reviewed. The operative wording is written by the site owner and their counsel before launch, and this notice is removed at that point.
Nothing on this page is a commitment, and no sentence here should be relied on until counsel has signed it off.
This is our responsible AI policy: what we will build, what we will not, and what we do when a system we built gets something wrong. It is written as commitments rather than principles, because a principle cannot be checked and a commitment can.
It applies to the agents we build for clients and to the agents we run ourselves. Where a client’s own framework is stricter, theirs applies.
Autonomy has a written boundary
A person stays accountable
Every action is recorded
Your data is not training data
Which model vendors we work with, and on what terms
What we will not build
Disclosure
When an agent gets it wrong
Human work we will not automate
How this policy changes
Contact
Related documents
- Security and governance architecture - the controls behind these commitments
- Privacy policy
- Terms of service
- Cookie policy
- Accessibility statement