01
Artificial intelligence
- Public-interest contribution
- Assisting analysis of large datasets, translation across languages, access to information and more efficient, responsive public services.
- Limits and risks
- AI systems can be inaccurate, biased or opaque, and may reproduce existing inequalities at scale. They are not a substitute for human judgement on consequential decisions.
- Required safeguards
- Human review of consequential decisions, independent evaluation in local contexts, documentation of limitations and clear routes to challenge outcomes.