• In November, the co-rapporteurs of the AI Act proposed the creation of an AI Office.

• Despite budgetary concerns, an AI Office or Agency would be an essential structure to regulate rapid AI advancements.

• Most importantly, an EU AI Agency could support SME innovators and advance digital sovereignty.

The proposal for an Artificial Intelligence Act aims to regulate the emerging field of AI systems in the Single Market. One of the core components of the proposal is the European AI Board which would facilitate the effective implementation of the regulation. Recently, the co-rapporteurs of the AI Act in the European Parliament proposed to replace the AI Board with an EU AI Office, an independent body with its own legal personality, funding, and staff.

DIGITAL SME explicitly welcomes the proposal for an EU AI Office or Agency. In our AI Open Letter, we advocated for the mandate of the AI Board to expand and go beyond compliance monitoring. The Letter was signed by 60 companies and business associations at the forefront of European innovation.

An AI Agency would constitute an appropriate governance structure that would be more apt to respond to the challenges of artificial intelligence and of the enforcement of the AI Act. These challenges can be broken down to two categories. The first one is the technical challenge of direct control of AI. This refers to the problem of ensuring that an AI system operates as it is intended and that regulatory safeguards are sufficient to prevent harmful misuses. The second one is the governance challenge of ensuring broad societal benefits of AI, i.e. of aligning the objectives of private actors and of their AI systems with the strategic goals of the European Union.

A centralised body that would enforce the AI Act could help us tackle both challenges. On the technical dimension, it is evident that AI is a field that is advancing rapidly. Moreover, the underlying algorithms of AI systems are complex, opaque and difficult to fully understand. A centralised body, composed of experts coming from the Member States, would be well-positioned to issue opinions, recommendations as well as binding decisions on issues of technical standards and regulatory sandboxes. Moreover, an AI Agency would engage in capacity-building with the national authorities through specialized training programmes on advanced AI systems.

On the social dimension, an AI Agency would be able to move away from the narrow mandate of the AI Board and engage in innovation monitoring. DIGITAL SME has repeatedly raised the issue of the financial burden that the AI Act would inflict on European SMEs. As a result of this burden, the EU is at risk to become entirely dependent on AI systems made elsewhere. An AI Agency could issue impact assessments on AI development and adoption in the Single Market. Moreover, it could assist SMEs in guiding the landscape of AI regulation and assist them in accessing funding and investment opportunities.

To fully realize its strategic purpose, the AI Agency would significantly benefit from a multi-stakeholder dialogue. The inclusion of academics, civil society, businesses and representatives of the SME community in the advisory board of the Agency would propel the adoption of innovation measures and the proliferation of best practices.

Throughout the history of the European Union, new regulatory bodies were created to respond to the challenges of new technologies. This was the idea behind the establishment of the European Atomic Energy Community, and of the European Union Agency for Cybersecurity (ENISA). Artificial intelligence is another disruptive technology that can bring a series of negative externalities in the day-to-day lives of citizens and businesses. This instance calls for a centralised, multi-stakeholder governance response to address the technical and societal challenges of AI. An EU AI Agency would be a critical investment towards a Europe fit for the digital age.