ECCC CYBERAI Funding 2027: Secure AI for European Cyber Operations

Applies to: technology providers, Cyber Hubs, CSIRTs, national and competent authorities, research organizations, NIS 2 entities and partners considering DIGITAL-ECCC-2027-DEPLOY-CYBER-11-CYBERAI.

The CYBERAI topic provides a €15 million envelope to develop and deploy secure, robust and trustworthy AI-powered systems for European cyber operations. The expected EU contribution is €3-5 million per project, the funding rate is 50%, and the indicative duration is 36 months. Proposals are due by 14 January 2027 at 17:00 CET and must include at least one official capability path, such as threat detection, CTI production, automated response, vulnerability management, self-healing, secure information sharing or protection of AI systems. A generic AI prototype without operational users, security controls and validation is not enough (official call document).

In Brief

• CYBERAI targets operational capacity for National and Cross-Border Cyber Hubs, CSIRTs, national authorities, NIS single points of contact and other relevant stakeholders.
• The topic covers both AI for cybersecurity and the cybersecurity of AI, including misuse, model risks, supply-chain security, robustness, trust and legal compliance.
• Activities must include at least one official path and should connect development, testing, validation, deployment and measurable operational use.
• Resulting systems, tools and services are expected to be available for licensing to Cyber Hubs, CSIRTs and relevant authorities under favorable market conditions.
• The topic is a Simple Grant with a 50% rate; no SME-specific 75% uplift applies.
• The two baseline KPIs measure benefiting entities and AI-powered capabilities provided to Cyber Hubs or national authorities.
• Article 12(5) security restrictions apply, which makes ownership, control, data location and technology dependencies material design issues.

Scope and Source Boundary

This guide explains the topic's operational scope, secure-AI expectations, stakeholder model, funding mechanics and evidence requirements. It does not determine the eligibility of a specific system or consortium and does not replace the official call document, the live Funding & Tenders record or applicable EU legislation.

CYBERAI is one of seven topics under DIGITAL-ECCC-2027-DEPLOY-CYBER-11. Its core purpose is to strengthen central cyber-operational capacity through AI and machine learning, including generative AI, while ensuring that the resulting technology is itself cybersecure and trustworthy (ECCC call overview).

CYBERAI at a Glance

The expected contribution is a planning range, not an automatic award or absolute cap. A different amount can be considered when duly justified, and the final grant may be lower than the amount requested.

CYBERAI or AI4SME?

Both topics concern AI-powered cybersecurity, but their primary outcomes differ.

Choose the topic from the intended operational result, not from the use of AI alone. A project may help SMEs and still fit CYBERAI if its primary deliverable is central operational capacity for Cyber Hubs or authorities; conversely, a user-friendly SME toolkit belongs more naturally under AI4SME.

Operational Capability Paths

The call requires activities to include at least one listed path. A focused proposal may combine several paths when their data, users and validation model form one coherent operating capability.

The topic also supports product security and cybersecurity by design or default in line with Cyber Resilience Act requirements.

Secure and Trustworthy AI Is Part of the Deliverable

The call explicitly requires attention to the security of AI itself, especially during learning, and to malicious use of AI. It also connects deployment to performance, robustness, trustworthiness, social acceptance, the AI Act, intellectual-property rules and the GDPR.

A proposal should address at least four linked layers:

1. AI asset layer: models, training and inference data, prompts, embeddings, pipelines, model registries and supporting infrastructure.
2. Threat layer: poisoning, evasion, adversarial input, model theft, data leakage, prompt injection, unsafe tool use, dependency compromise and malicious repurposing where applicable.
3. Control layer: provenance, access control, isolation, integrity checks, validation, monitoring, confidence thresholds, human authorization, rollback and incident handling.
4. Assurance layer: test cases, acceptance criteria, independent review, reproducibility, limitations, audit logs and evidence for certification or standardization.

ENISA's Multilayer Framework for Good Cybersecurity Practices for AI and Cybersecurity of AI and Standardisation provide official reference points cited by the call. The project should still adapt controls to its own architecture, operating context and risk appetite.

Data, CTI and Information-Sharing Design

AI cyber operations depend on data from authorities, Cyber Hubs, CSIRTs, NIS entities and technical sensors. Data volume alone does not demonstrate quality or legality.

A defensible data model should define:

• source, ownership, authorization and permitted purpose for each dataset;
• classification, sensitivity, personal-data and intellectual-property constraints;
• normalization, deduplication, labeling, quality and confidence controls;
• anonymization or pseudonymization before cross-organization sharing;
• access segregation, geographic location, retention and deletion;
• feedback from analysts and downstream CTI consumers;
• handling of false or adversarial intelligence;
• export formats and interoperability with operational platforms.

For original European CTI feeds, the proposal should explain what makes the intelligence original, who can consume it, how confidence is expressed and how the service remains sustainable after funding.

Deployment, Licensing and Strategic Dependencies

The call expects systems, tools and services to be made available for licensing under favorable market conditions to National or Cross-Border Cyber Hubs, CSIRTs, competent authorities and other relevant authorities. This makes exploitation and licensing architecture part of the project design rather than an end-of-project afterthought.

Define early:

• foreground and background intellectual property;
• licensing model, eligible users and support obligations;
• deployment options for sensitive environments;
• maintenance, model updates and vulnerability handling;
• portability and exit arrangements;
• dependencies on non-eligible suppliers, models, data or cloud infrastructure;
• controls against foreign influence and indirect control.

The call stresses that Cyber Hubs must be protected from dependencies and vulnerabilities because of their operational role and sensitive information. In well-justified cases, access to EuroHPC high-performance computing infrastructure may be granted; this is a possibility, not an entitlement, so the proposal needs a viable compute plan without assuming approval.

Who Can Participate and How to Build the Consortium

Targeted stakeholders include technology providers, Cyber Hub operators, research and academia, cybersecurity entities, the public sector, NIS 2 entities, private-sector organizations and other actors supporting secure-AI deployment.

The common rules impose no minimum consortium composition for CYBERAI, and a single applicant may submit where all topic and capacity conditions are satisfied. A typical end-to-end partnership may include:

Every partner should own a measurable part of the operational result. A consortium with impressive names but no data rights, deployment environment or adoption commitment will remain weak.

Funding and Cost Model

CYBERAI is a Simple Grant with a 50% funding rate. The expected EU contribution is €3-5 million per project, with an indicative duration of 36 months.

The budget-based grant reimburses eligible actual costs with the unit-cost and flat-rate elements defined in the call. Planning should account for:

• indirect costs at 7% of applicable eligible direct costs;
• depreciation and full cost for listed equipment under the topic conditions;
• the no-profit rule and possible reductions for non-compliance;
• co-financing, cash flow and financial capacity for each beneficiary;
• eligible-country restrictions for work, subcontracting and controlled entities.

Compute, data engineering, secure infrastructure and validation can create substantial costs. Each amount should trace to workloads, environments, work packages and measurable outputs. Do not rely on a broad "AI platform" budget line without usage assumptions and cost ownership.

Expected Outcomes and Evidence Model

The expected outcomes include deployed AI capabilities for Cyber Hubs and authorities, validated novel tools, better collaboration, automated CTI processes, original European CTI feeds, advanced secure AI for NIS sectors and contributions to standardization or certification.

Validation should cover relevant conditions, not only laboratory accuracy. For an operational tool, evidence may need to include performance under load, adversarial behavior, degraded inputs, model drift, human response and recovery from incorrect automation.

KPI Contract

The topic defines two mandatory indicators:

1. Number of entities benefiting from development, testing, validation, deployment or uptake of AI-powered cybersecurity technologies, tools and services.
2. Number of AI-powered cybersecurity technologies and capabilities provided to National or Cross-Border Cyber Hubs and national authorities encompassing Cyber Hubs and CSIRTs.

Optional indicators cover AI services for rapid detection and decision-making, automated threat detection and response, AI-IoT-smart-grid risk-reduction tools, original CTI feeds, cybersecurity-by-design tools and protection of AI solutions.

For each applicable KPI, define baseline, target, unit, data source, collection frequency and responsible partner. If an indicator does not apply because the related activity or outcome is outside the project, state and justify that boundary. Additional KPIs should capture genuine gaps such as detection quality, decision time, deployment availability, analyst workload, robustness or safe-response performance.

A Practical Work-Package Model

This is a planning pattern, not a mandatory ECCC template.

The common rules also require a dissemination and exploitation deliverable in the first six months and yearly deliverables on relevant KPIs and project outputs.

How Evaluators Will Read the Proposal

For CYBERAI, the award subcriteria on overcoming lack of market finance and on environmental sustainability or European Green Deal effects are explicitly not applicable. The digital technology supply-chain subcriterion remains applicable.

Passing the thresholds does not guarantee funding. Proposals are ranked within the available topic budget, and legal, financial, exclusion and security checks continue during grant preparation.

Eligibility, Security and Application Package

Beneficiaries and affiliated entities must generally be eligible legal entities established in EU Member States or in Norway, Iceland or Liechtenstein. Article 12(5) restrictions apply to participation in every role: entities must be established in and controlled from eligible countries, while project activities and subcontracted work must take place there.

The application consists of online Part A, technical Part B limited to 70 pages, and required ownership and control declarations. Submission is exclusively electronic through the Funding & Tenders Portal.

Readiness Checklist

1. Confirm that central cyber-operational capacity is the primary result and select the correct topic.
2. Choose at least one official capability path and map every work package to it.
3. Secure operating-user commitment, data rights and a representative validation environment.
4. Define AI assets, threats, controls, assurance evidence and acceptable automation boundaries.
5. Document GDPR, intellectual-property, AI Act and CRA considerations where applicable.
6. Plan how results will be licensed to Cyber Hubs, CSIRTs or relevant authorities under favorable conditions.
7. Address supplier, model, cloud, data and control dependencies under the security restrictions.
8. Include both mandatory KPIs and justify every optional or non-applicable indicator.
9. Build a credible 50% co-financing plan and evidence the cost of compute, data, pilots and assurance.
10. Use the live portal templates, keep Part B within 70 pages and submit before the deadline.

Frequently Asked Questions

Conclusion

CYBERAI is designed for projects that can turn secure AI into validated European cyber-operational capability. A strong proposal joins authoritative users, governed data, proven technology, adversarial testing, measurable operational improvement and a realistic licensing path.

We can help you define security requirements, evidence models, data governance and risk controls through our Virtual CISO service; the Cyber Check-up gives each partner a first documented measure of its posture. Final scope, eligibility, legal, cost and submission decisions must remain anchored to the official call document, applicable legislation and live Funding & Tenders record.

Guides in the ECCC 2027 Series

• Digital Europe Cybersecurity Call 2027: €96 Million Across Seven ECCC Topics
• ECCC Cybersecurity Call 2027: Eligibility and Application Guide
• ECCC AI4SME Funding 2027: Guide for SMEs and Cybersecurity Providers
• ECCC EULEG Funding 2027: From Cybersecurity Law to Shared Capability
• ECCC COORDPREP 2027: Funding for Coordinated Cyber Preparedness Testing
• ECCC NCC Network Funding 2027: How National Coordination Centres Support Cyber Ecosystems
• ECCC Regional Cable Hubs 2027: Funding for Cross-Border Undersea Cable Security
• ECCC Dual-Use Cybersecurity Funding 2027: From Civilian-Defence Cooperation to Deployment

Official Sources

• ECCC - Official call document, version 1.0
• ECCC - Call overview and CYBERAI summary
• ENISA - Multilayer Framework for Good Cybersecurity Practices for AI
• ENISA - Cybersecurity of AI and Standardisation
• European Commission - Funding & Tenders call search