We're delighted to announce our first pilot grant cohort, where we are supporting four organisations in sub-Saharan Africa, the Arab region, and the Brazilian Amazon with €2.73M in funding.
These four organisations are each tackling cultural preservation from different angles, covering ecosystem infrastructure, community governance, accountability, and cultural knowledge. Together they cover the full range of what public-interest AI actually requires: the conditions that make ethical data collection possible, the infrastructure that makes cultural knowledge legible to AI systems, the sovereignty frameworks that keep communities in control throughout, and the tools to hold it all to account.
Meet our grantees
Masakhane, Kenya
Masakhane is the most prominent grassroots African language AI hub on the continent, working across 50+ languages spoken by over 1 billion people. They've spent years building the connective tissue for African language AI from the ground up with communities who have never seen themselves in these systems. Their model of community-led development is exactly what the field needs more of.
We're supporting their continued institutional independence and their expansion into domain-specific datasets for health, farming, and education. Much of their work begins at the source: gathering and structuring data in languages that AI has largely never seen, building the linguistic infrastructure others can build on.
Institute for Worldmaking (IWM), Lebanon
Arab cultural knowledge is largely absent from AI systems. The Institute for Worldmaking is building public-interest cultural infrastructure to change that — digitising and structuring Arab cultural history and contemporary practice into public, multilingual, machine-readable knowledge across universities and cultural institutions in the region, as well as Arab artists and researchers in diaspora.
Our funding supports two AI-ready cultural databases, an auditable, retrieval-based AI agent for research and education, and governance frameworks that enable Arab communities to shape how their knowledge enters AI on their own terms. Without a structured, consented, machine-readable infrastructure, cultural knowledge simply doesn't enter AI systems. That's the gap IWM is building to close
Portal sem Porteiras, Brasil
Indigenous languages are being absorbed into AI training pipelines without community consent, often using data scraped without permission. Portal sem Porteiras is building a different model. Working with communities in the Brazilian Amazon and Cerrado, they've established governance-first principles: communities define which AI tools are built and maintain control over their use.
Our funding supports their offline-first, locally-hosted tools that keep data ownership within the territory. By keeping communities in control at every stage, from what gets recorded to how models are trained and deployed, PSP shows what it looks like to navigate the full process without ever losing sovereignty over the results.
African Internet Rights Alliance, Kenya
AI systems increasingly shape rights and access to services across Africa, but effective accountability guardrails are largely absent. The African Internet Rights Alliance is building it — developing scalable, context-appropriate AI audit tools designed for reuse by regulators, civil society organisations, and global policy actors. AI systems are already making consequential decisions across Africa. AIRA is building the tools to scrutinise them.
The cohort is intentionally diverse by design. What Masakhane learns about building African language infrastructure informs how PSP thinks about data sovereignty. What AIRA develops in audit methodology is available to everyone. What IWM builds in cultural knowledge infrastructure points toward what's possible elsewhere. Each organisation brings their expertise, and together we aim to create something greater than the sum of its parts
We can’t wait to see where they go.