AI Governance
We use AI where it helps with drafting and development, with human judgement responsible for what goes live.
Principles
- Human in the loop. People remain accountable for consequential decisions and for content that is published.
- Disclosure. We consider when AI use should be made clear to the people affected.
- Data minimisation. We pass only the fields a model needs. We aim to share only information needed for the task and review a tool’s data handling before use.
- Explainability. Information generated with AI needs a source and a way for a person to check it.
Bias & fairness
We consider whether AI-assisted outputs risk excluding people or misrepresenting information, and review them before use.
Vendors
Tools should be chosen with attention to data handling, privacy and the information shared with them.
Logging & audit
We review AI use and maintain appropriate records where a project requires them.