Public AI is first of all an idea: that AI is and can be a form of public infrastructure like highways, parks, libraries, water, electricity, and the Internet itself. The movement for public AI, or Public AI (capitalized), is a global movement committed to building public AI. It is not a single organization but a union of independent projects and contributors working toward a shared vision.

We coordinate around specific projects, including:

• Public inference - the Public AI Inference Utility is a pilot making public AI models more accessible and demonstrating real user demand for public AI. The Swiss Public Inference Utility is a nation-level approach to this.
• Frontier models - Fully open-source models like Apertus, Olmo, SEA-LION, and Salamandra as well as projects like Airbus for AI promoting multilateral collaboration toward a public-private frontier-scale model.
• Movement building - the Public AI Network is a community of builders and advocates supporting one another in the movement towards more public AI.
• Public access in libraries – Drop-in AI makerspaces across public libraries.
• Benchmarks – Educational resources on AI benchmarks and evaluations; platforms like Weval for collaborative benchmarking.
• Data - Data flywheels for gathering usage data to improve public AI, and an AI Commons of well-provenanced public knowledge for training and grounding AI systems.

We welcome engineers, researchers, policymakers, designers, and community members of all stripes and backgrounds. Explore repositories, open issues, or bring new projects aligned with the Public AI mission.

Public AI is built by all of us.