Summary Under the proposed Cloud and AI Development Act (CADA), Grand Challenge 6 is dedicated to developing cooperative European industrial AI models that enable large-scale collaboration across strategic sectors without exposing commercially sensitive data. As defined in Annex I(6) of the proposal, this initiative prioritizes advanced confidentiality-preserving technologiesβ€”including federated learning, secure execution environments, encryption-based processing, and anonymisationβ€”to allow competitors and partners to pool data securely. Target sectors explicitly include aerospace, pharmaceutics, cybersecurity, mobility, autonomous vehicles and drones, energy, and defence. This challenge operates under the Cloud and AI Leadership Initiatives (Article 6) and complements Operational Objective 5 (Article 4), aiming to overcome the "data silo" bottleneck that currently hinders European industrial AI competitiveness.

Detail

The Cloud and AI Development Act (CADA), as proposed in COM(2026) 502 final, establishes the Cloud and AI Leadership Initiatives to bridge the gap between the Union's advanced research capabilities and their sustainable exploitation in the industrial market. A central mechanism of these initiatives is the designation of "grand challenges"β€”large-scale, cross-sectoral projects designed to address major technological and industrial hurdles of strategic relevance for the Union.

Grand Challenge 6, explicitly detailed in Annex I(6) of the proposal, is titled "Cooperative European Industrial Models." Its primary objective is to foster the development of AI models and systems for strategic sectors by enabling collaboration at a European industrial scale without exposing commercially sensitive data between participants. This addresses a critical market failure: the reluctance of European companies to share proprietary datasets due to fears of intellectual property leakage, competitive disadvantage, or regulatory non-compliance.

Technical Mechanisms: Confidentiality-Preserving Technologies

To achieve the goal of secure collaboration, Grand Challenge 6 mandates a specific focus on advanced confidentiality-preserving technologies. The proposal outlines a suite of mechanisms designed to allow algorithms to learn from distributed data without the data itself ever leaving the owner's secure environment. As set out in Annex I(6), these mechanisms include:

  • Federated and distributed training approaches: These methods involve bringing the algorithms to the data rather than transferring the data to a central location. This ensures that raw datasets remain under the control of the original data holder.
  • Secure execution environments: These provide isolated computing environments where data can be processed securely, preventing unauthorized access or extraction during the training phase.
  • Encryption-based processing: This ensures that data remains encrypted and unreadable even while being computed upon, protecting it from interception or misuse.
  • Anonymisation and pseudonymisation techniques: These methods strip identifiable or sensitive markers from datasets before they are used in collaborative models, reducing privacy risks.
  • Access compartmentalisation: This limits data visibility to specific, authorized functions or personnel, ensuring that no single entity has full visibility into the entire dataset.
  • Protections against extraction: These measures are designed to prevent the reverse-engineering or extraction of commercially sensitive information from the trained models themselves, ensuring that the output does not inadvertently reveal the input data.

Target Strategic Sectors

The proposal identifies specific strategic sectors that stand to benefit most from these cooperative models. These sectors are characterized by high value, complex global supply chains, and strict data privacy or security requirements. Annex I(6) explicitly lists the following sectors as priority areas for Grand Challenge 6:

  • Aerospace: Enabling collaboration on design and manufacturing data without compromising proprietary engineering secrets.
  • Pharmaceutics: Facilitating joint drug discovery and clinical trial analysis while protecting patient data and intellectual property.
  • Cybersecurity: Allowing for the sharing of threat intelligence and attack patterns to build more robust defensive models without revealing sensitive network configurations.
  • Mobility: Including autonomous vehicles and drones, where collaboration on sensor data and navigation algorithms is crucial for safety and efficiency.
  • Energy: Supporting the optimization of grid management and renewable energy integration through shared data.
  • Defence: Enabling secure collaboration on sensitive defence technologies and operational data.

Legal Framework and Implementation

Grand Challenge 6 is implemented under the broader framework of the Cloud and AI Leadership Initiatives, governed by Article 6 of the CADA proposal. Article 6(2) states that the operational objectives of these initiatives shall be implemented through "large-scale, cross-sectoral initiatives addressing major technological and industrial challenges of strategic relevance for the Union," as indicated in Annex I.

This challenge is closely linked to Operational Objective 5 (Industrial AI), which is detailed in Article 4(5) of the proposal. While Grand Challenge 5 focuses on accelerating the development and uptake of sector-specific AI models, Grand Challenge 6 specifically targets the cooperative and confidential aspects of multi-party model training. Article 4(5)(c) explicitly mandates the initiative to "enable secure large-scale data pooling for collaborative AI training through technologies enhancing privacy and preserving confidentiality."

Furthermore, Article 6(4) empowers the Commission to adopt delegated acts to amend Annex I in a manner consistent with the objectives of the Cloud and AI Leadership Initiatives. This ensures that the definition of cooperative models and the specific technologies required can be updated to reflect rapid advancements in privacy-enhancing technologies (PETs) and market developments.

The implementation of these initiatives is supported by Article 9, which provides for the allocation of AI computing resources to priority projects. Projects designated as "frontier AI priority projects" under Article 8 may receive matching computing resources from the Union's share of European high-performance computing (EuroHPC) capacity, ensuring that the computational demands of large-scale cooperative training are met.

What this means for you

For CTOs, data architects, R&D directors, and SMEs operating in the listed strategic sectors, Grand Challenge 6 represents a paradigm shift in how industrial AI can be developed and deployed. It moves the industry from a model of data hoarding to one of secure, value-driven collaboration.

1. Breaking the Data Silo Barrier

Historically, industries like pharmaceuticals and aerospace have operated in isolated data silos due to intense IP protection concerns and competitive pressures. Grand Challenge 6 provides a regulatory and technical framework for breaking these silos. Organizations can now explore collaborative AI initiatives with competitors, suppliers, or research partners without the legal and operational risks associated with sharing raw, proprietary datasets. This could unlock significant value in areas like joint drug discovery or shared supply chain optimization.

2. Strategic Investment in Privacy-Enhancing Technologies (PETs)

To participate in or benefit from these initiatives, organizations must invest in and adopt Privacy-Enhancing Technologies (PETs). If your current architecture does not support federated learning, secure multi-party computation (MPC), homomorphic encryption, or secure enclaves, you may be excluded from future collaborative AI projects funded or recognized under CADA. Architects should begin evaluating their infrastructure's compatibility with these technologies immediately. The proposal signals that the future of European industrial AI is built on these foundations.

3. Access to Shared Compute Resources

CADA aims to significantly increase access to high-performance computing (HPC) resources for AI projects. By aligning your R&D strategy with the objectives of Grand Challenge 6, your organization may gain priority access to EU-funded compute capacity. Article 9 explicitly mentions the allocation of AI computing resources from EuroHPC capacity to support frontier AI priority projects. Collaborative projects that demonstrate the use of confidentiality-preserving technologies are likely to be strong candidates for such support.

4. New Opportunities for SMEs

The proposal places a strong emphasis on the inclusion of SMEs. Article 33 requires Member States to monitor procurement of innovation and actively support SME participation. Grand Challenge 6 projects, being large-scale and cross-sectoral, will likely require specialized components that only niche players can provide. If your SME offers expertise in encryption, secure execution environments, or specific industrial domain knowledge, you are well-positioned to bid for subcontracts or to lead specific work packages within these cooperative models.

5. Standardization and Market Leadership

As these cooperative models become mainstream, the need for common technical standards will be critical. Early adopters who help define the technical standards for secure execution environments, federated training protocols, and data compartmentalisation in these sectors will gain a significant competitive advantage. Participation in the standardization process, potentially facilitated by the Centres for AI established under Article 5, could position your organization as a market leader in the European industrial AI ecosystem.

Common misconceptions

Misconception 1: Grand Challenge 6 requires sharing raw data. Correction: No. The entire premise of Annex I(6) is to enable collaboration without exposing commercially sensitive data. The focus is on techniques like federated learning where the model learns from local data without the data ever leaving the local environment. The data remains under the control of the original owner.

Misconception 2: This only applies to large corporations. Correction: While the initiatives are large-scale and cross-sectoral, CADA explicitly aims to foster the ecosystem for SMEs. Article 4(5) and Article 33 highlight the importance of SMEs in innovation procurement and industrial AI uptake. SMEs can participate as specialized providers of PETs, domain-specific models, or secure infrastructure components.

Misconception 3: It replaces existing industrial AI efforts. Correction: Grand Challenge 6 complements Grand Challenge 5 (Industrial AI). While GC5 focuses on accelerating the development and deployment of sector-specific AI models, GC6 focuses specifically on the cooperative and confidential aspects of multi-party model training. They are synergistic, with GC6 providing the secure framework necessary for GC5's broader deployment goals.

Misconception 4: The technical standards are already finalized. Correction: CADA is a proposal. The specific technical criteria for "secure execution environments" and "encryption-based processing" will be further defined through delegated acts and implementing measures. Article 6(4) empowers the Commission to amend Annex I to reflect technological developments. The framework sets the direction, but the detailed technical specifications are still being developed.

Related

This is general information about a draft EU regulation, not legal advice.