A high bar for calling something 'good'
'AI for Good' is easy to say and hard to earn. We use it to mean AI that demonstrably improves human outcomes, with people in control.
What we mean by it
For us, AI for Good is not a category of project — it is a test that a project has to pass. Does it improve access, learning, care or community capability for real people? Can that improvement be observed and, ideally, measured? Do the people affected have a say and a way to opt out? If the answer to those questions is no, it is not AI for Good, however impressive the technology.
Where we look
Accessibility
Communication access, accessible information and services that reach Deaf, disabled and multilingual communities.
Education
AI literacy and learning support that build understanding and independence rather than dependence.
Healthcare & care access
Improving access to information and navigation of care — never replacing clinical judgement or collecting health data here.
Community capability
Helping community organisations and social-impact builders use responsible AI well.
Responsible deployment
Rolling out carefully, monitoring for harm, and being ready to stop or change course.
Human oversight
Keeping accountable people in the loop for decisions that affect lives.
Measuring impact honestly
We prefer a small, verified improvement to a large, unverifiable claim. Where we describe outcomes, we describe how we know. Where a project is illustrative rather than active, we label it clearly.
Illustrative use cases
- Illustrative example
Accessible information
Helping public information — notices, guides, essential services — reach people across languages, reading levels and formats, with human review of every output. The test: can the people it is meant for actually understand and use it?
- Illustrative example
Community AI literacy
Plain-language learning, built with community organisations, that helps people understand what AI can and cannot do, when it is being used on them, and what their rights are. The test: do participants leave more confident and more independent, not more dependent?
- Illustrative example
Service navigation
Helping people find and understand services they are entitled to — without collecting sensitive data, and with a clear human pathway when a question matters too much to leave to software. The test: does it widen access without creating a new gatekeeper?