AI Ethics Checklist

Format: Conduct an ethics review before launching an AI feature.

Checklist:

Transparency:
□ Do users know they are interacting with AI?
□ Do users know AI can make mistakes?
□ Can users opt out of the AI feature?

Fairness:
□ Does AI perform consistently across different groups?
□ Is the training data biased?
□ Could someone be harmed by AI's mistakes?

Privacy:
□ Will users' inputs be used to train models?
□ Do users know how their data is being used?
□ Can users delete their own data?

Safety:
□ Can AI be manipulated to produce harmful output?
□ Could someone use AI to gain unauthorized information?
□ Is AI's failure mode safe? (Better to refuse than give a harmful answer)