Summary For AI developers, the most important definitions in the proposed Cloud and AI Development Act (CADA) would be ‘AI system’, ‘frontier AI’ and ‘AI agent’. All three sit in Article 2 and, as proposed, would shape eligibility for Union support: recognition as a frontier AI priority project, access to matched AI computing resources, and support for large-scale AI agent platforms. Getting these distinctions right is how you find the support mechanisms CADA would offer.

Detail

CADA is a proposal (COM(2026) 502 final), not yet in force. It would introduce definitions that frame its support measures. For developers, architects and CTOs, three Article 2 terms are pivotal — they would act as gateways to specific opportunities.

AI system Article 2(3) defines an ‘AI system’ by reference to Article 3, point (1), of the Artificial Intelligence Act (Regulation (EU) 2024/1689). This cross-reference keeps CADA consistent with the AI Act. In practice, if your product is already an AI system under the AI Act, it would fall within CADA’s ecosystem measures using the same definition.

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." This term matters because it identifies technologies eligible for the strongest support. Projects recognised as frontier AI priority projects under the criteria in Article 8 would, under Article 9, benefit from allocated AI computing resources — with the Union committing to "at least match the AI computing resources contributed by Member States" to such projects, to the extent capacity is available within the Union’s share of European high performance computing access time.

AI agent 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." Notably, this includes a "coordinated set of AI systems," so multi-agent architectures can fall within scope. The proposal places real weight here: under operational objective 6 of Article 4, the Cloud and AI Leadership Initiatives would support the development, deployment and orchestration of advanced AI agents at scale.

Together, these definitions would create a structured way to target resources — funding, high-performance computing, and secure platforms for autonomous AI agents — to where the proposal judges they are most needed.

What this means for you

For CTOs, architects and SMEs, these definitions would have direct planning implications.

Eligibility for funding and support If you develop AI models or systems that "approach, reach or exceed the current state of the art," you may meet the ‘frontier AI’ definition. That could open a path to recognition as a frontier AI priority project under Article 8 and, in turn, allocated AI computing resources under Article 9 — including the Union’s commitment to at least match Member State contributions, within the limits of available capacity.

Platform development and deployment The ‘AI agent’ definition signals the proposal’s focus on autonomous, tool-using systems. If you build or deploy AI agents, operational objective 6 of Article 4 explicitly aims to support agent development, deployment and orchestration at scale — relevant for those working on orchestration frameworks, multi-agent systems, or autonomous execution.

Regulatory alignment Because Article 2(3) borrows the AI Act’s definition of ‘AI system’, you would not be navigating two divergent definitions. If you already track AI Act compliance, you are likely within CADA’s definitional scope, letting you focus on the support measures — for example the Experience and Acceleration Centres for AI (‘Centres for AI’).

Strategic positioning If your system is advanced but not at the ‘frontier’, you may still benefit from other measures under the Cloud and AI Leadership Initiatives, such as those for industrial AI or physical AI. Classifying your technology accurately against these terms helps you find the right support track.

Common misconceptions

Misconception 1: CADA defines AI systems independently of the AI Act. Article 2(3) explicitly references the AI Act’s definition, so there is no separate CADA test for ‘AI system’ — the same definition applies across both.

Misconception 2: Only the largest companies can be ‘frontier AI’ developers. The Article 2(4) definition turns on capability (approaching, reaching, or exceeding the state of the art), not organisation size. SMEs and research institutions developing cutting-edge models could qualify on the merits.

Misconception 3: AI agents are a niche concept. A dedicated Article 2(5) definition, paired with operational objective 6 of Article 4, signals that autonomous, tool-using systems are a central focus of the proposal’s support measures — not a side topic.

Misconception 4: CADA imposes heavy new compliance duties on developers. CADA is framed primarily as a support-and-investment instrument for the cloud and AI ecosystem. It references the AI Act for definitions but, as proposed, its developer-facing measures are largely about funding, compute access and platforms rather than new AI safety obligations (which remain the AI Act’s domain).

Official sources

Related

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