# Responsible Ai Checklist

## Purpose and accountability
- [ ] The intended use and prohibited uses are documented.
- [ ] An accountable human or team is named.
- [ ] A simpler non-Ai solution was considered.

## People and fairness
- [ ] Affected people and possible harms are identified.
- [ ] Performance is checked across relevant groups and conditions.
- [ ] Users have an escalation, correction or appeal route where appropriate.

## Privacy and data
- [ ] Only necessary data is collected and retained.
- [ ] Consent, rights, access and security are addressed.
- [ ] Training and retrieved content cannot expose restricted information.

## Safety, security and misuse
- [ ] Prompt injection, data poisoning, unauthorized actions and misuse are considered.
- [ ] Tools and permissions follow least-privilege principles.
- [ ] Important actions require validation or confirmation.

## Transparency and oversight
- [ ] Users know when they are interacting with Ai.
- [ ] Important limitations and uncertainty are communicated.
- [ ] Human reviewers have authority, information, time and training.

## Operations
- [ ] Quality, incidents, drift and harmful outcomes are monitored.
- [ ] A rollback or shutdown method exists.
- [ ] The review date and responsible owner are recorded.