One shared system. Three connected ambitions.

The strategic goal

CEREBRA-AI aims to establish Europe’s reference framework for autonomous AI. Its shared intelligence layer is intended to help independently developed models, tools and specialist capabilities work together, learn from outcomes and improve as one system.

This strategic direction brings the project’s research areas into a common task-to-outcome loop. The formal project title remains “A unified framework for continual, explainable and embodied General-Purpose Artificial Intelligence”.

Scientific ambition: improvement that transfers

The scientific questions concern how systems reason, adapt and learn through interaction. The project will investigate whether composition and experience improve complete-task outcomes, whether earlier capabilities are retained, and whether gains remain useful in another setting.

Evaluation must separate the contribution of the framework from improvements in the models underneath it. Strong baselines, controlled resources and independent reproduction are central to that distinction.

Technological ambition: one execution loop

The technical direction is a runtime that receives a task and its limits, selects compatible capabilities, supervises execution, checks outcomes and records the evidence. Learning and release processes then determine which changes are ready for future tasks.

Shared descriptions of context, capability, task, outcome and change are intended to make this loop reusable across applications. People retain authority over purpose, permissions, approvals and intervention.

Commercial ambition: a foundation partners can build on

The project aims to create a reusable framework around which partners can develop specialist capabilities, sector products, integration and operational services. Practical value will be assessed through useful outcomes and the effort required to integrate, operate and update systems.

These are routes for exploitation and sustainability. Licensing, ownership, maintenance and service arrangements will be established for the relevant assets as the programme develops.

A European operating layer

The longer-term ambition is an interoperable layer connecting compute, models and data with sector applications. Compatible providers and deployment environments should remain replaceable, while shared interfaces and evidence support continuity.

European reference status is an ambition to earn through research, adoption and stewardship. Wider deployment beyond the project will require further validation and sustainable operating arrangements.

How the programme is organised

Work package	Focus
WP1	Project management, data governance, ethics and coordination.
WP2	Hybrid neuro-symbolic reasoning and the core architecture.
WP3	Continual and evolutionary learning, and specialist-model orchestration.
WP4	Embodied and relational learning, interaction and environment interfaces.
WP5	Explainability, trustworthiness, fairness and compliance support.
WP6	Integration, sector adaptation and pilot validation.
WP7	Open science, dissemination, exploitation and ecosystem engagement.
WP8	Ethics requirements.

Project facts

Item	Details
Acronym	CEREBRA-AI
Programme	Horizon Europe
Action	Research and Innovation Action
Grant agreement	101299046
Topic	HORIZON-CL4-2025-04-DIGITAL-EMERGING-07
Topic title	Enhanced Learning Strategies for General Purpose AI: Advancing GenAI4EU
Start	01/09/2026
End	31/08/2030
Duration	48 months
EU contribution	€14,997,790
Total eligible project costs	€14,997,790
Coordinator	NEURALIO AI IKE, Greece
Granting authority	European Health and Digital Executive Agency (HaDEA)