EU Frontier AI Initiative: Developing frontier AI solutions that are safe and computationally efficient within Apply AI (RIA)

The EU Frontier AI Initiative focuses on developing safe and computationally efficient frontier AI solutions under the Horizon Europe program, specifically in the DIGITAL sector. The primary funding mechanism is a grant, categorized as a HORIZON Research and Innovation Action, with a total budget of €44 million for this specific call, titled HORIZON-CL4-2027-04-DIGITAL-EMERGING-11. Eligible applicants include legal entities from EU Member States, Iceland, Norway, Canada, Israel, Republic of Korea, New Zealand, Switzerland, and the United Kingdom, while third countries can participate if they are associated with Horizon Europe during 2026-2027.

The initiative aims to strengthen European capabilities in developing frontier AI models by ensuring computational efficiency, enhancing safety, and promoting research in AI methodologies. It encourages consortium participation from various sectors, including academia and industry, to foster collaboration and innovation. The expected project maturity encompasses research, development, and demonstration phases, emphasizing the validation of AI systems through practical implementations.

The application process is single-stage, with a planned opening date of November 17, 2026, and a deadline of March 18, 2027. Only one project is expected to be funded, indicating a highly competitive selection process, likely resulting in a success rate below 10%. Funding covers eligible research and innovation costs, with no mandatory co-funding requirement for applicants.

The initiative intends to develop advanced AI models with multimodal capabilities while addressing the challenges of energy efficiency and safety, including algorithms for robustness against adversarial attacks and enhancing interpretability. Proposals must also establish strong links with initiatives like the Resource for AI Science in Europe and engage with industry partners to support project outcomes.

This funding opportunity represents a significant investment in the future of AI in Europe, aligning with broader digital strategies and enhancing the region's autonomy in AI technology development through rigorous research and innovation efforts.

This EU funding opportunity focuses on the EU Frontier AI Initiative: Developing frontier AI solutions that are safe and computationally efficient within Apply AI (RIA). It falls under the Horizon Europe (HORIZON) program, specifically the DIGITAL (HORIZON-CL4-2027-04) call. The type of action is HORIZON-RIA (Research and Innovation Actions), utilizing a HORIZON Action Grant Budget-Based [HORIZON-AG] Model Grant Agreement. The deadline model is single-stage, with a planned opening date of November 17, 2026, and a deadline of March 18, 2027, at 17:00:00 Brussels time.

The expected outcomes of this funding opportunity are: Strengthened European capabilities in the development of frontier AI models, improved computational efficiency of frontier AI models, resulting in reduced computational costs, and enhanced safety of advanced AI systems based on frontier AI models through the development and implementation of safe-by-design principles and/or AI agents acting as safety evaluators.

The scope of this opportunity is to advance developments of frontier AI models towards highest-level performance, while ensuring energy efficiency, addressing computational constraints, and strengthening safety. The approach is twofold: advancing the AI field through the development and training of a frontier AI model with state-of-the-art performance, multimodal capabilities, and optimization for agentic AI capabilities, and supporting research on comprehensive methods to reduce the computational demands of frontier AI models and ensure their safety, including technical methodologies like automated testing and interpretability.

Key research areas include compression and distillation techniques for reducing the complexity of large AI models, innovations in AI architectures to lower computational demands, and algorithmic approaches to minimize computational load during pre-training, post-training, and inference. Addressing misalignment of large AI models, enhancing robustness against adversarial attacks, jailbreaks, and backdoors, and advancing AI models transparency and interpretability are also crucial.

This initiative contributes to the EU Frontier AI initiative and requires strong links with the Resource for AI Science in Europe (RAISE). Activities are expected to involve the European AI research community and attract top AI talent. Proposals should incorporate mechanisms for assessing progress, including KPIs, benchmarking, and progress monitoring. Results should be shared through the AI-on-demand platform. Projects should link to AI Factories and Data Labs and establish links with European companies developing frontier AI models. Cohesion activities with the European Partnership on AI, data, and robotics (ADRA) and the CSA HORIZON-CL4-2025-03-HUMAN-18: GenAI4EU central Hub are also expected.

Admissibility conditions include proposal page limits and layout as described in Annex A and Annex E of the Horizon Europe Work Programme General Annexes and Part B of the Application Form. Eligible countries are described in Annex B of the Work Programme General Annexes. Participation is limited to legal entities established in Member States, Iceland, Norway, Canada, Israel, Republic of Korea, New Zealand, Switzerland, and the United Kingdom. Entities controlled by non-eligible countries may not participate unless guarantees are provided.

Financial and operational capacity and exclusion criteria are described in Annex C of the Work Programme General Annexes. Evaluation and award criteria, scoring, and thresholds are described in Annex D. Submission and evaluation processes are described in Annex F and the Online Manual. The indicative timeline for evaluation and grant agreement is in Annex F. Legal and financial set-up of grants is described in Annex G. Specific conditions are described in the specific topic of the Work Programme.

Application forms are available in the Submission System, with standard forms for HE RIA and IA. Evaluation form templates will be used with adaptations, and standard evaluation forms for HE RIA and IA are available. Guidance is provided in the HE Programme Guide, HE MGA, and call-specific instructions, including an Ownership Control Declaration Annex.

Additional documents include the HE Main Work Programme 2026-2027 (General Introduction, Digital, Industry and Space, and General Annexes), HE Programme Guide, HE Framework Programme 2021/695, HE Specific Programme Decision 2021/764, EU Financial Regulation 2024/2509, Rules for Legal Entity Validation, LEAR Appointment and Financial Capacity Assessment, EU Grants AGA, Funding & Tenders Portal Online Manual, Terms and Conditions, and Privacy Statement.

The budget overview for 2027 includes various topics: HORIZON-CL4-2027-04-DATA-03 (35,000,000 EUR), HORIZON-CL4-2027-04-DATA-08 (38,000,000 EUR), HORIZON-CL4-2027-04-DATA-09 (39,000,000 EUR), HORIZON-CL4-2027-04-DIGITAL-EMERGING-04 (42,000,000 EUR), HORIZON-CL4-2027-04-DIGITAL-EMERGING-05 (18,000,000 EUR), HORIZON-CL4-2027-04-DIGITAL-EMERGING-06 (3,000,000 EUR), HORIZON-CL4-2027-04-DIGITAL-EMERGING-10 (4,000,000 EUR), HORIZON-CL4-2027-04-DIGITAL-EMERGING-11 (44,000,000 EUR), HORIZON-CL4-2027-04-HUMAN-01 (39,000,000 EUR), HORIZON-CL4-2027-04-HUMAN-02 (3,000,000 EUR), and HORIZON-CL4-2027-04-HUMAN-07 (7,000,000 EUR). All topics follow a single-stage submission process with a planned opening date of November 17, 2026, and a deadline of March 18, 2027.

Partner search announcements are available for collaboration. The submission system is planned to open on the date stated on the topic header. General FAQs and support are available through the Funding & Tenders Portal.

This funding opportunity is designed to bolster the EU's capabilities in AI by supporting the development of advanced, efficient, and safe frontier AI models. It encourages research into reducing computational demands and ensuring the safety of these models, fostering a sovereign European AI ecosystem. The initiative aims to bring together top AI talent and resources to drive innovation and address critical challenges in the field, aligning with the EU's broader digital strategy and ensuring that AI development is both cutting-edge and ethically sound.

Eligible Applicant Types: The eligible applicant types are legal entities established in Member States, Iceland, Norway, Canada, Israel, Republic of Korea, New Zealand, Switzerland, and the United Kingdom. Additionally, entities from third countries that may become associated with Horizon Europe during 2026 and 2027 may be eligible if their country is identified as eligible at the time of application submission. There are also restrictions on entities controlled by non-eligible countries or those assessed as high-risk suppliers of mobile network communication equipment.

Funding Type: The funding type is primarily a grant, specifically a HORIZON Research and Innovation Action (HORIZON-RIA) and HORIZON Innovation Actions (HORIZON-IA) and HORIZON Coordination and Support Actions (HORIZON-CSA) under the Horizon Europe program. The type of MGA is HORIZON Action Grant Budget-Based [HORIZON-AG].

Consortium Requirement: The opportunity appears to require a consortium, as it is geared towards larger projects and encourages partner searches. The text mentions partner requests and collaboration.

Beneficiary Scope (Geographic Eligibility): The geographic eligibility includes EU Member States, Iceland, Norway, Canada, Israel, Republic of Korea, New Zealand, Switzerland, and the United Kingdom. It also includes third countries that may become associated with Horizon Europe during 2026 and 2027, contingent on their eligibility status at the time of application.

Target Sector: The program targets the digital sector, specifically focusing on artificial intelligence (AI), data, and robotics. It emphasizes the development of frontier AI models, computational efficiency, and AI safety.

Mentioned Countries: Iceland, Norway, Canada, Israel, Republic of Korea, New Zealand, Switzerland, United Kingdom, EU Member States.

Project Stage: The project stage targets research, development, and innovation, with an emphasis on advancing existing AI models and developing new methodologies. It spans from development and training to implementation and safety assurance.

Funding Amount: The funding amounts vary depending on the specific topic within the call, ranging from €3,000,000 to €44,000,000. For example, HORIZON-CL4-2027-04-DIGITAL-EMERGING-11 has a budget of €44,000,000, while HORIZON-CL4-2027-04-DIGITAL-EMERGING-06 has a budget of €3,000,000. The HORIZON-CL4-2027-04-HUMAN-01 topic has a budget of €39,000,000 with individual grants ranging from €4,800,000 to €5,600,000.

Application Type: The application type is an open call with a single-stage submission process.

Nature of Support: The beneficiaries will receive money in the form of grants to support their research and innovation activities.

Application Stages: The application process involves a single stage.

Success Rates: The success rates are not explicitly mentioned, but the indicative number of grants for each topic provides some insight. For instance, HORIZON-CL4-2027-04-DIGITAL-EMERGING-11 aims for one grant, while HORIZON-CL4-2027-04-HUMAN-01 aims for eight grants.

Co-funding Requirement: The information does not explicitly state a co-funding requirement.

Summary: This Horizon Europe call, under the DIGITAL program, focuses on the EU Frontier AI Initiative, aiming to bolster European capabilities in developing safe, computationally efficient, and high-performance AI models. The call invites proposals for Research and Innovation Actions (RIA), Innovation Actions (IA) and Coordination and Support Actions (CSA) that advance the AI field, improve computational efficiency, and enhance the safety of AI systems. Eligible applicants include legal entities from EU Member States, associated countries, and certain third countries, with specific conditions to safeguard the Union’s strategic assets. The funding amounts vary by topic, and the application process involves a single-stage submission. The initiative seeks to foster collaboration within the European AI research community and link projects with AI Factories, Data Labs, and relevant European companies.