HORIZON-CL4-2027-04-DIGITAL-EMERGING-11

EU Frontier AI Initiative: Developing Sovereign, Safe and Computationally Efficient Frontier AI Models

Frontier AI
Safe by Design
Agentic AI
High-Performance Compute

Programme

EU Frontier AI Initiative: Developing Sovereign, Safe and Computationally Efficient Frontier AI Models

Note📣 Call Overview: HORIZON-CL4-2027-04-DIGITAL-EMERGING-11
Field Details Field Details Field Details
Status

Forthcoming

Opening Date

2026-11-17

Closing Date

2027-03-18

Target TRL Foundational Validation (TRL 2 to 4) Type Action

Research and Innovation Action

Type Grant

Action Grant Budget-Based

Submission Model Single-stage Geographic Target

European AI Research Community (Strict China Restrictions)

Budget per Bid

~€44.00M

Expected Slots 1 slot available Total Allocation

€44.00M

Note: The deadline closes strictly at 17:00:00 Brussels time. This is an elite, high-value single-winner call (only 1 project will be funded across Europe). Critical Geopolitical Eligibility Guardrail: Legal entities established in China are strictly ineligible to participate in this destination.

🎁 Expected Outcome by the EU

ImportantExpected Outcomes: HORIZON-CL4-2027-04-DIGITAL-EMERGING-11
  • 🇪🇺 Sovereign Strategic Autonomy: Significantly strengthen European capabilities and achieve excellence in the development of independent, state-of-the-art European frontier models.
  • Environmental & Financial Efficiency: Radically improve the computational efficiency of frontier architectures, lowering the energy footprint and reducing training and inference costs to lower barriers to entry.
  • 🛡️ Hardcoded Safe-by-Design Foundations: Enhance the core security of advanced AI via automated “safe-by-design” principles and/or specialized AI agents functioning as autonomous safety evaluators.

🏗️ Scope Summary

WarningScope: HORIZON-CL4-2027-04-DIGITAL-EMERGING-11

The initiative demands a dual-track research methodology combining frontier model generation with cutting-edge socio-technical safety instrumentation:

🧠 1. Core Model Development & Optimization

  • Multimodal State-of-the-Art Training: Advancing a native frontier AI model optimized for sophisticated “agentic AI capabilities” (including autonomous problem-solving, advanced reasoning, and tool use).
  • Algorithmic Weight Compression: Applying distillation, compression, and novel architecture innovations to minimize raw computational load across pre-training, post-training, and inference phases.

⚖️ 2. Advanced Alignment & Safety Evaluation

  • Sophisticated Misbehavior Mitigation: Developing technical methodologies to detect and stop high-order misalignments, including “alignment faking”, encoded reasoning within chain-of-thought (CoT) processes, and the “reward hacking of human oversight”.
  • Structural Robustness: Engineering defenses against adversarial injections, jailbreaks, backdoors, and the systemic risks unique to embedding frontier models inside autonomous agent frameworks.

🔗 Required Infrastructure & Ecosystem Integration: Consortia must establish operational links with the Resource for AI Science in Europe (RAISE), AI Factories, and Data Labs. Projects must feed communicable results into the AI-on-demand platform and coordinate cohesion tasks alongside the European Partnership on AI, Data, and Robotics (ADRA) and the GenAI4EU Central Hub.

🛣️ Strategic Alignment

This Research and Innovation Action (RIA) represents the pinnacle of the EU’s long-term “AI Continent” ambition mapped out in the Apply AI Strategy. Moving from basic formulation (TRL 2) to laboratory-validated proof of concept (TRL 4), the call addresses economic security as outlined in the Draghi Report and Competitiveness Compass.

Our research framework directly interfaces with this foundational layer, anchoring systemic transparency onto sovereign, next-generation European compute infrastructures.

TipSCVCS’ Technical Value Proposition

We provide the underlying governance, RegTech capabilities, and architectural frameworks necessary to anchor an AI Environmental/Reputation Governance Harness.

By introducing our socio-technical architecture, we supply the concrete audit layers needed to track and evaluate safety metrics. Our framework operates as an independent governance layer alongside the call’s required AI safety evaluators—verifying algorithmic transparency, tracking the model’s energy footprint, and auditing alignment profiles without disrupting the core high-performance computing pipeline.


🤝 SCVCS’ Core Areas of Collaboration

We are utilizing our research frameworks to establish structural bridges with partners targeting the following call mandates:

  1. 🔏 Socio-Technical Alignment & Safe-by-Design Architecture: Translating regulatory principles into mathematical and architectural constraints. We match our frameworks with the IEEE International Roadmap for Devices and Systems (IRDS) Environment, Safety, Health, and Sustainability (ESHS) roadmaps to counter adversarial vulnerabilities and verify that chain-of-thought (CoT) reasoning models remain fully auditable.

  2. 🌊 Computational Efficiency Tracking & Carbon Audits: Designing data-logging layers compatible with the International Data Spaces architecture to benchmark and audit real-time model compression. This provides a transparent ledger for verifying the ecological efficiency of large-scale pre-training runs.

  3. 🏫 Consortium Cohesion & Talent Hubs: Supporting the integration requirements connecting foundational research networks with industrial application verticals, ensuring sovereign European AI capabilities seamlessly transfer into validated global value chain services.

Note👷 Consortium Engagement

Because this is a hyper-competitive, single-winner call (1 funded project), we are strictly aligning with top-tier, institutional bids:

  • 🇪🇺 Major European AI Research Labs & Coordinators: Core scientific institutions leading large-scale consortium assembly who require an elite socio-technical partner to manage safety KPIs, audit frameworks, or alignment governance structures.
  • 🏭 Supercomputing Centers & Frontier AI Enterprises: Infrastructure providers and European AI firms looking to formalize their integration alongside RAISE or AI Factories while embedding rigorous safety-by-design compliance.

📬 Action Item: To review joint deployment strategies or exchange Participant Identification Code (PIC) profiles, please reach out via our Contact Page.


Official Reference: HORIZON-CL4-2027-04-DIGITAL-EMERGING-11 - EU Funding & Tenders Portal