Community Ownership
Affected communities need continuous mechanisms to shape and critique the AI systems being built for them.
- Summary
- Related Resources
- Projects
Community Ownership
Affected communities need continuous mechanisms to shape and critique the AI systems being built for them.

What concrete veto power or governance mechanism does the community hold over your tool once it is deployed in the wild?
Why it matters
People affected by an AI system rarely get a say in how it is built. Decisions about what data to collect, which outcomes to optimise for and what counts as an acceptable error are usually made by a small team, far from the people who live with the results.
Community governance gives affected people a standing role in those decisions — not a one-off survey at launch, but a mechanism that keeps running for as long as the system does. Civic tech is built in the name of the public; governance is how the public gets to hold it to that promise.
Principles
- Involve people early. Bring the community in while the problem is still being defined, not after the model is trained.
- Share real power. Advice that can be ignored is not governance. Agree up front which decisions the community can block or change.
- Make it continuous. Systems drift, data changes and new harms appear. Review on a schedule, not only when something goes wrong.
- Lower the cost of taking part. Pay people for their time, meet where they already are, and explain the system without jargon.
- Close the loop. Report back on what changed because of community input — and what did not, and why.
What it looks like in practice
- A community advisory group that reviews major changes before they ship
- A public channel for reporting harm, with a committed response time
- Plain-language reports on how the system is used and where it fails
- A clear way to opt out, and to have personal data removed
- A sunset plan: who decides when the system should be retired
- Who is affected by this system but has no say in how it works?
- What would it take for them to stop it if it caused harm?
- When did we last change something because a community member asked?
Governance takes time and money. Budget for it from the start, the same way you would for hosting or security — retrofitting it after launch costs more.
- Algorithmic Justice League — research and advocacy on the harms of AI systems, led by affected communities
- Ada Lovelace Institute — research on public participation in data and AI governance
- Data & Society — research on the social implications of data-centric technologies
- All topics — related topics on accountability and transparency
- Seedscore — an impact assessment tool that builds community and stakeholder questions into project planning