AI Development
Responsible AI
The rules we follow on every project, so you can use AI with confidence and explain it to others.
Principles
Six rules
People stay in charge
Every system has a named reviewer. For decisions that affect individuals, a person makes the call and the system supports.
Honest about limits
We test on your material, report where the system is weak, and do not promise accuracy we cannot show.
Careful with data
Before work starts we agree in writing where data is stored, who can see it, how long it is kept and how it may be used.
A record you can read
Each system keeps a log of what it did and why, so you can answer questions from auditors, boards and the public.
Permission for voices and likenesses
Voice and likeness features are used only with written consent from the person concerned.
Plain disclosure
Where audiences see AI-produced material, it is disclosed, as with the AI-generated imagery noted sitewide on this site.
In practice
How the rules show up in a project
Each project begins with a short risk review. We list the harms that could occur, who would notice, and what stops them. The list has owners and dates.
After launch, Model Operations & Oversight keeps testing the system against fresh material and reports the results to you.
- Risk review before build
- Reviewer named for each system
- Regular accuracy checks
- Clear disclosure to audiences
Questions about risk?
We are happy to walk your legal, security or leadership teams through our process.