CEREBRA-AI will deliver a sovereign, trustworthy-by-design general-purpose AI platform built around four pillars: (1) a hybrid neuro-symbolic reasoning core for verifiable multi-step inference; (2) a continual-learning orchestrator that composes specialist models, detects drift, and balances accuracy, latency, and energy; (3) an embodied and relational interface that grounds knowledge through interaction with robots and high-fidelity digital twins; and (4) a transversal trust framework embedding explainability, provenance, fairness auditing, uncertainty handling, and AI Act–aligned compliance. The architecture is operationalised via an integration substrate that connects to European data spaces, orchestrates retrieval-augmented and agentic workflows, and provides conformity tooling for interoperable, evidence-backed deployments from edge to cloud. Impact will be demonstrated across sectoral pilots in healthcare, mobility, energy, environment, and manufacturing, targeting ≥20% performance gains while maturing assets to TRL-5 and reporting safety and trust metrics. The project commits to open assets, open-access publications, and large-scale training, with deposits on AI-on-Demand and sustained engagement with European ecosystems and standards bodies to drive uptake. A rigorous risk and exploitation plan underpins sustainability: resource-aware design and access to compute mitigate data and scaling bottlenecks; a clear KER portfolio, tech-transfer agreements, and open-source stewardship ensure post-project continuity and market traction. By coupling reasoning, lifelong adaptation, embodiment, and built-in compliance in a single EU-aligned stack, CEREBRA-AI advances the state of the art and strengthens Europe’s digital autonomy with a practical, auditable path to trustworthy GPAI.

Project funded under the call HORIZON-CL4-2025-04, grant 101299046