The AI Alliance has unveiled Project Tapestry, a new open-source initiative aimed at enabling collaborative and sovereign development of advanced artificial intelligence models.

The project will create a distributed platform for training large-scale AI systems across global participants, allowing institutions, companies, and governments to contribute to shared models while retaining control over their own data and applications.

Project Tapestry is designed to address growing concerns over the concentration of AI development within a small number of companies and regions. By enabling federated collaboration, the platform seeks to lower barriers related to compute, data access, and expertise that often limit independent AI efforts.

“As AI is fast becoming part of the common infrastructure, there is a need for foundation models to be open so as to enable sovereignty and cultural diversity,” said Yann LeCun, who will join the AI Alliance as Chief Science Advisor. “Project Tapestry is an ambitious effort to bring that model to AI — to create the conditions for open, distributed progress on systems of real capability.”

The initiative will support the creation of a shared global base model that participants can extend into localized or sector-specific versions aligned with their own priorities, regulations, and values.

“Project Tapestry is designed to overcome that constraint through federated collaboration,” said Christopher Nguyen, chief architect of the project. “The idea is simple but powerful: build a shared global base openly, then enable each participant to extend it in ways they fully own and control.”

The AI Alliance said the platform will be governed by a global board and supported through its nonprofit research organization, with further technical development to be shaped through international collaboration workshops beginning in May.