Summary The proposed Cloud and AI Development Act (CADA) does not prescribe a fixed administrative form for "frontier AI priority projects." Instead, Article 8 establishes three cumulative criteria that applications must satisfy: the project must be pioneering, undertaken by an eligible consortium involving at least three Member States, and demonstrate pooled computing resources. Applicants must respond to open calls for expressions of interest by providing evidence that maps directly to these statutory thresholds. The regulation defines the qualifications for recognition, not a rigid checklist of documents; therefore, the documentation required is whatever evidence is necessary to prove compliance with Article 8(a) through (c).

Detail

Under the proposed Cloud and AI Development Act (CADA), the designation of a "frontier AI priority project" serves as a strategic mechanism to scale up essential breakthroughs in artificial intelligence and reduce dependencies on third-country technologies. The process is governed primarily by Article 8 of the proposal, which sets out the criteria for recognition by the Commission.

It is crucial to understand that the regulation defines the qualifications for a project, not a rigid checklist of documents. However, to satisfy the Commission, applicants must generate and submit evidence that maps directly to the three cumulative criteria listed in Article 8(a) through (c). The text explicitly states that the Commission may recognize projects "provided that the following criteria are fulfilled." Consequently, the documentation burden is on the applicant to construct a case that satisfies these specific legal conditions.

1. The Legal Framework: Article 8 Criteria

Article 8 states that the Commission may recognize projects as frontier AI priority projects if they support "grand challenge 3" (Frontier AI) set out in Annex I, provided that specific conditions are met. The text of Article 8 requires that:

  • Article 8(a): Pioneering Focus. The project must be a "pioneering project, focused on the support and scaling-up of frontier AI technologies."
  • Article 8(b): Eligible Entity and Cross-Border Participation. The project must be "undertaken by a European digital infrastructure consortium established pursuant Decision (EU) 2022/2481 or another legal entity eligible for funding under Union law and it involves the participation of at least three Member States."
  • Article 8(c): Pooled Resources. The "participating Member States pool computing time and other relevant resources to support the implementation of the designated project."

These criteria are cumulative; failure to meet any one of them precludes recognition. The documentation submitted must therefore address each point explicitly.

2. Evidence Required for Article 8(a): Pioneering Frontier AI

To satisfy the first criterion, the application documentation must demonstrate that the project is not merely incremental but represents a strategic advancement in frontier AI. Article 2(4) defines "frontier AI" as "AI models or AI systems built upon such models that can perform a wide variety of tasks and that approach, reach or exceed the current state of the art."

Documentation should include:

  • Technical Roadmap and Grand Challenge Alignment: A detailed description of the architectural design and development of next-generation multimodal models. This must align with Grand Challenge 3 in Annex I, which focuses on "superior performance in advanced reasoning, cross-modal understanding and agentic capabilities." The application should explicitly reference how the project addresses the "grand challenge" of developing "next-generation multimodal models and systems."
  • State-of-the-Art Comparison: Evidence showing how the proposed models will exceed current benchmarks. This may involve preliminary research data, theoretical models, or prototypes that demonstrate "novel approaches to model efficiency, cognitive modelling, and alternative computational structures."
  • Strategic Asset Declaration: A statement explaining how the project contributes to the Union's strategic autonomy. Recital 16 notes that "fostering the development of frontier AI technologies as strategic assets should also reduce current dependencies on third-country technologies." The application should articulate this strategic value.

3. Evidence Required for Article 8(b): Eligibility and Participation

This criterion addresses the legal structure and geographic scope of the applicant. The documentation must prove both the legal eligibility of the entity and the breadth of its European participation.

Documentation should include:

  • Proof of Legal Status: If applying through a European Digital Infrastructure Consortium (EDIC), provide the registration documents establishing the consortium under Decision (EU) 2022/2481. If using "another legal entity eligible for funding under Union law," provide incorporation documents and evidence of eligibility for EU funding instruments (e.g., Horizon Europe or the Digital Europe Programme). The text of Article 8(b) is specific: the entity must be an EDIC or another eligible legal entity.
  • Member State Participation Letters: Formal letters of commitment or participation from the governments or designated national authorities of at least three Member States. These documents must confirm their involvement in the project governance and execution. Article 8(b) explicitly requires "the participation of at least three Member States." A project involving only two Member States would fail this criterion.
  • Governance Structure: Organizational charts showing how the consortium or legal entity is structured to facilitate cross-border collaboration, ensuring that the project is not dominated by a single national interest but serves the Union's collective goals.

4. Evidence Required for Article 8(c): Pooled Computing Resources

The third criterion requires concrete evidence of resource commitment. This is critical because the Union will match these contributions with its own EuroHPC capacity, as per Article 9. Article 9(2) states that "The Union shall at least match the AI computing resources contributed by Member States to frontier AI priority projects."

Documentation should include:

  • Compute Commitment Agreements: Legally binding agreements or memoranda of understanding from the participating Member States detailing the specific amount of AI computing resources (e.g., FLOPs, GPU hours) they will contribute. The text of Article 8(c) requires that "participating Member States pool computing time and other relevant resources."
  • Resource Pooling Mechanism: A technical and administrative plan describing how these resources will be pooled and managed. This should outline the governance of the compute pool, access protocols for the project team, and how resources will be allocated to ensure efficient use.
  • Complementarity Plan: Evidence that the pooled resources are specifically targeted to support the implementation of the frontier AI project, distinguishing them from general-purpose compute allocations. Article 9(1) notes that resources are allocated "within the limits of available capacity," so the application must demonstrate the feasibility of the pooling arrangement.

5. Responding to Open Calls

Article 8 specifies that projects are selected through "open calls for expression of interest." Therefore, the documentation must be tailored to the specific requirements of each call. The Commission will issue these calls, and the application package must adhere to the procedural guidelines set out in the call documents.

The regulation itself does not list a "detailed file list" of administrative forms. Instead, it sets the criteria that the expression of interest must satisfy. Applicants should monitor the Commission's announcements for these calls, as the specific submission portals, formatting requirements, and deadlines will be defined in the call documentation, not in the text of CADA itself. The core requirement remains: the expression of interest must provide sufficient evidence to satisfy the three cumulative criteria of Article 8(a), (b), and (c).

What this means for you

As a cloud service provider, data centre operator, or research institution looking to participate in or support a frontier AI priority project, your role is primarily evidentiary and technical. You are not applying directly as the project lead (unless you are part of the eligible consortium), but your infrastructure is the backbone of the Article 8(c) requirement.

Actionable Steps:

  1. Prepare Resource Certificates: Ensure you have the capability to issue precise certificates of compute capacity (in FLOPs or equivalent metrics) that can be formally committed by a Member State. Vague estimates will not suffice to prove the "pooling" required by Article 8(c).
  2. Align with Grand Challenge 3: Review Annex I, Grand Challenge 3. Your technical documentation should explicitly map your infrastructure's capabilities to the needs of "next-generation multimodal models" and "agentic capabilities."
  3. Verify Consortium Eligibility: If you are joining a consortium, verify that it is either an EDIC established under Decision (EU) 2022/2481 or an entity eligible for Union funding. Ensure that at least three Member States are formally onboarded before the expression of interest is submitted.
  4. Draft Pooling Agreements: Work with national authorities to draft clear, legally sound agreements that define the "pooling" of resources. Ambiguity here could lead to rejection under Article 8(c).
  5. Monitor Open Calls: Since Article 8 relies on "open calls for expression of interest," stay alert for Commission announcements. The specific submission format will be defined in those calls, not in the regulation.

Common misconceptions

Misconception 1: There is a standard application form in CADA. Reality: CADA Article 8 sets criteria, not a form. The "documentation" is the evidence you construct to prove you meet the criteria. The actual submission process will be defined by the specific open call issued by the Commission.

Misconception 2: Any AI project can apply. Reality: Only projects that are "pioneering" and focused on "frontier AI" (as defined in Article 2(4)) are eligible. Incremental improvements to existing models do not qualify. The project must also support "Grand Challenge 3" from Annex I.

Misconception 3: Two Member States are enough. Reality: Article 8(b) explicitly requires the participation of "at least three Member States." A bilateral project between two countries does not meet the threshold for a frontier AI priority project under this framework.

Misconception 4: Private companies can apply directly. Reality: The application must be undertaken by a European digital infrastructure consortium (EDIC) or another legal entity eligible for Union funding. Private companies must participate through these eligible structures, not as standalone applicants.

Misconception 5: The regulation lists every document needed. Reality: The text of Article 8 sets the criteria (pioneering nature, eligible entity, pooled resources) but does not enumerate a specific list of files. The applicant must determine what evidence is necessary to prove compliance with these criteria.

Official sources

Related

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