DeepRAP: Deep Reasoning, Abstraction & Planning towards trustworthy Cognitive AI Systems

The DeepRAP Challenge supports research and innovation in trustworthy cognitive AI, advancing scientific state-of-the-art and positioning Europe as a leader in human-centric, safe and explainable AI. Proposals should deliver cognitive AI systems capable of deep reasoning, abstraction, and planning, integrating provable trustworthiness mechanisms with alignment to ethical, legal, and societal standards (including the AI Act). Projects are expected to develop interoperable, application-driven AI systems, demonstrate these in real-world and simulated scenarios, create new methods for AI evaluation and certification, follow FAIR data principles, and collaborate within a portfolio to ensure coverage of capabilities, technological approaches, and use cases. The ultimate goal is to go beyond current deep learning and symbolic AI paradigms, exploring innovative methods inspired by neuroscience, biology, philosophy, and other fields, to achieve breakthroughs in abstraction, contextualisation, causality, and human-like cognitive functions.

Check your compatibility

Agency	HORIZON (EU)
Region	EU
Country	N/A
Legal form	Company, NGO, Government organization, Private person, Other
Start date	22 Jul 2026
End date	28 Oct 2026
Max funding	Up to 4 million EUR per project
Budget	96 million EUR for 2026 (Pathfinder Challenges budget, including this and two other actions)
Coverage	Minimum requested contribution per project: 500,000 EUR; maximum: 4,000,000 EUR

Grant application conditions

Applicant structure	Single legal entities or consortia of legal entities established in an EU Member State or Associated Country. Consortia of two: independent legal entities from two different Member States or Associated Countries. Consortia of three or more: at least three independent legal entities, each from a different country (minimum one in a Member State, two others from different Member States or Associated Countries).
Eligible applicants	Universities, research organisations, SMEs, start-ups, natural persons. Single beneficiary projects: mid-cap and larger companies are NOT permitted.
Page limit	Max. 30 format A4 pages for Sections 1–3 of Part B. Extra pages will not be considered.
Legal and financial set-up	Lump Sum Model Grant Agreement applies.
Portfolio coverage	Cognitive Function Capability: Projects should address at least one – reasoning, abstraction, planning. Technological Approach: Diversity expected (e.g., neuro-symbolic AI, deep learning, reinforcement learning, interdisciplinary frameworks). Use Cases: Proposals should apply to real-world domains such as industry, mobility, civil security, scientific discovery, health, cybersecurity, justice or human-robot interaction.
Documentation	https://ec.europa.eu/info/funding-tenders/opportunities/portal/screen/opportunities/topic-details/HORIZON-EIC-PATHFINDERCHALLENGES-01-03

Eligible expenses

Evaluation criteria

Business exclusions

Reporting

Application process

Application requirements