Summary Yes, frontier AI priority projects under the proposed Cloud and AI Development Act (CADA) canβ€”and are explicitly encouraged toβ€”develop AI agents. The proposal defines "AI agents" in Article 2(5) and identifies "agentic capabilities" as a core strategic objective in Annex I, Grand Challenge 3. However, to qualify for the "frontier AI priority project" designation and access associated computing resources, an initiative must strictly satisfy the structural criteria of Article 8, including formation as a European digital infrastructure consortium (EDIC) and participation by at least three Member States.

Detail

The proposed Cloud and AI Development Act (CADA), COM(2026) 502 final, establishes a targeted framework to bolster the EU's technological sovereignty in artificial intelligence. A central pillar of this framework is the "Cloud and AI Leadership Initiatives," which includes a specific mechanism to designate and support "frontier AI priority projects." For technology leaders and architects, a critical question is whether the development of autonomous AI agents falls within the scope of this support. The answer is affirmative, provided the project adheres to the specific definitions and structural requirements laid out in the proposal.

The Definition of AI Agents in CADA

CADA provides a precise legal definition for AI agents, distinguishing them from static AI systems. Article 2(5) defines an "AI agent" as:

"an AI system or a coordinated set of AI systems, that can perceive and act upon their environment, with a degree of autonomy, using tools as needed to achieve specific goals and adapt to changing inputs and contexts."

This definition is not merely descriptive; it establishes the functional baseline for what the EU considers a strategic asset. By codifying the ability to "perceive," "act," and "adapt" with autonomy, the proposal signals that systems capable of autonomous execution are a priority for Union-level intervention. This definition serves as the foundation for the operational objectives of the Cloud and AI Leadership Initiatives, ensuring that funding and support mechanisms are aligned with the development of these dynamic, tool-using systems.

Agentic Capabilities as a Strategic Focus

The strategic intent to support AI agents is further reinforced in the annexes of the proposal. Annex I, which outlines the "Grand Challenges" for the Cloud and AI Leadership Initiatives, dedicates Grand Challenge 3 to "Frontier AI." The text of this challenge explicitly lists "agentic capabilities" alongside "advanced reasoning" and "cross-modal understanding" as key areas for pioneering novel capabilities.

The proposal states that the focus of Grand Challenge 3 will be on:

"architectural design and development of next-generation multimodal models and systems that push the boundaries of current algorithmic capabilities for achieving superior performance in advanced reasoning, cross-modal understanding and agentic capabilities."

This confirms that the development of frontier multimodal systems with autonomous, agentic behaviors is not only permitted but is a targeted area for EU support. The proposal envisions these agents as critical for applications ranging from scientific discovery to complex data interpretation, where autonomous management and planning are required. Therefore, a project focused on developing such agents aligns directly with the "grand challenges" the Act seeks to address.

The Article 8 Criteria for Priority Projects

While the technology (AI agents) is a priority, the project must meet rigorous administrative and structural criteria to be recognized as a "frontier AI priority project" under Article 8. The Commission may recognize projects selected through open calls for expression of interest that support Grand Challenge 3, but only if they fulfill three cumulative conditions:

  1. Pioneering Focus: The project must be a "pioneering project, focused on the support and scaling-up of frontier AI technologies." This ensures that the designation is reserved for initiatives pushing the state of the art, rather than incremental improvements.
  2. Consortium Structure: 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." This requirement mandates a collaborative, cross-border legal structure, preventing single-entity dominance and ensuring broad European participation.
  3. Multi-State Participation: The project must "involve the participation of at least three Member States." This condition is designed to foster integration across the Union's digital single market and ensure that the benefits of the technology are distributed geographically.

Consequently, a single SME, a national research institute, or even a large multinational corporation cannot independently apply for status as a frontier AI priority project. The model requires a collaborative, cross-border approach. This structural hurdle is significant: it means that to develop frontier AI agents under this specific CADA pathway, organizations must form or join a consortium that spans at least three EU Member States.

Access to Computing Resources

Once a project meets the Article 8 criteria and is designated as a frontier AI priority project, it gains access to significant computational resources under Article 9. The proposal stipulates that:

"The Union and the Member States shall ensure that sufficient AI computing resources from their compute capacities are allocated to support the development of frontier AI priority projects that fulfil the criteria set out in Article 8, within the limits of available capacity."

Crucially, Article 9(2) establishes a matching mechanism:

"The Union shall at least match the AI computing resources contributed by Member States to frontier AI priority projects to the extent that sufficient AI computing capacity is available within the Union's share of European high performance computing access time."

This provision is a critical incentive for developers of resource-intensive frontier models and complex agent orchestration frameworks. By matching national contributions with Union-level EuroHPC capacity, CADA aims to reduce the compute gap that often hinders European AI development compared to global competitors.

What this means for you

For CTOs, research directors, and architects evaluating the development of AI agents, the CADA framework offers a clear but demanding pathway to support.

  • Validate the Technology: Ensure your project explicitly targets the definition in Article 2(5). Your system must demonstrate the ability to perceive, act, and adapt autonomously using tools. If your project is a static model without these agentic features, it may not qualify under the "agentic capabilities" focus of Grand Challenge 3.
  • Build a Cross-Border Consortium: You cannot apply alone. To qualify as a priority project, you must partner with entities in at least two other Member States and structure the project under a European digital infrastructure consortium (EDIC) or an equivalent Union-eligible legal entity. Start networking now to identify partners who can meet the "three Member States" threshold.
  • Align with Grand Challenge 3: When drafting proposals, explicitly map your technical roadmap to the "agentic capabilities" and "advanced reasoning" objectives listed in Annex I. Demonstrate how your agents will push the boundaries of current algorithmic capabilities, rather than just applying existing models to new data.
  • Leverage the Matching Mechanism: If you secure priority status, you gain access to matched EuroHPC resources. Plan your compute architecture to leverage this matching, ensuring you have the national contributions ready to trigger the Union-level allocation.

Common misconceptions

"Any AI agent project qualifies for frontier support." Incorrect. While AI agents are a priority, only projects that meet the strict Article 8 criteriaβ€”specifically the multi-state consortium requirement and the focus on scaling frontier technologiesβ€”can be designated as priority projects. Standard AI agent development that does not involve a cross-border consortium or does not push the state of the art will not qualify.

"Frontier AI priority projects are only for large hyperscalers." While large players will likely lead these consortia, the requirement for a European digital infrastructure consortium allows SMEs, research institutions, and startups to participate as essential partners. The key is the collaborative, cross-border structure, not just the size of a single entity.

"CADA defines 'frontier AI' solely by parameter count." CADA does not set a rigid parameter threshold in the main text for priority projects. Instead, it relies on the functional definition in Article 2(4) ("approach, reach or exceed the current state of the art") and the strategic focus on capabilities like agentic behavior and multimodal reasoning outlined in Annex I. A project with fewer parameters but superior agentic autonomy could theoretically qualify if it meets the "pioneering" criteria.

Official sources

Related

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