The Rockefeller Foundation and OpenAI Foundation have launched a $50 million fund for organisations helping city governments use AI in public services. Grants will range from $500,000 to $2 million, and applicants can use whichever AI models they choose, rather than being steered towards OpenAI.
OpenAI Watch analysis
What happened
The Cities Deliver initiative is inviting applications from eligible non-profits and civil-society organisations. Each project must have a city-government partner, serve a charitable purpose and include a meaningful frontier AI component. The application window opened on 14 October and closes on 29 November 2026.
The programme focuses on public health, economic mobility, environmental resilience and civic engagement. Projects relating to voting or elections are excluded. The AI Insider reports that the fund aims to reach up to 100 million city residents.
The foundations also plan follow-on grants for projects in Cali, Cape Town and Rio de Janeiro. These include expanding an AI flood-alert system in Rio and extending Dengue.AI outbreak forecasting from Cali to Cartagena.
Why it matters
This is a substantial attempt to move civic AI beyond pilots and conference slides, with grants large enough to support real deployments. Requiring a local-government partner may help connect projects to public services, while allowing applicants to choose their models avoids making the fund a disguised product funnel.
The catch is that a grant announcement is not evidence of successful public services. The programme’s value will depend on what cities build, how residents benefit and whether the systems work reliably in the messy conditions of local government.
Our read
The open model choice is the detail to watch: it makes the initiative more credible as public-interest funding, and gives applicants room to choose tools that fit their needs. The application criteria are also concrete enough for eligible organisations to assess whether they have a viable city partner and project before the November deadline.
What to watch
- Which organisations and cities receive grants, and how funding is distributed across the four focus areas.
- Whether the named follow-on projects publish measurable results.
- How applicants demonstrate that their AI component is meaningful and useful to residents.
Discussion spark: Should public-interest AI grants require cities to prove measurable benefits before projects scale, or would that make it too difficult to fund worthwhile experiments?
Sources and evidence
OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.