EU AI Act Takes Effect Today, But Belgium Has No Supervisor

The European Union’s AI Act enters its next phase today, introducing mandatory transparency rules for artificial intelligence systems across all member states. Yet as the new obligations take effect, Belgium remains without an officially designated national supervisory authority to enforce them, leaving a significant gap in compliance oversight.

Under the new rules, companies must clearly disclose when users are interacting with AI systems rather than humans, and AI-generated or manipulated content must be visibly labelled. According to VRT NWS, “if a company puts an AI image online instead of a real photo, from now on it must be noted that the image was made or edited with AI.” Similarly, customers chatting with an online store’s customer service must know whether they are speaking with a chatbot or a human.

What the New Transparency Rules Mean

The AI Act (Regulation (EU) 2024/1689), the world’s first comprehensive legal framework for artificial intelligence, was formally adopted in May 2024 and entered into force on 1 August 2024. It is being implemented in phases, with the European Commission confirming that from 2 August 2026, its AI Office will begin enforcing the rules alongside national authorities.

The new transparency obligations under Article 50 require AI systems to tell users when they are interacting with AI, and AI-generated or manipulated content must carry machine-readable marks for easier detection. The European Commission notes that these measures are “intended to reduce deception and manipulation and help people make informed choices.” Companies violating the transparency rules face fines of up to €15 million or 3% of global annual turnover.

Belgium’s Missing Supervisor

Every EU member state was required to designate a national market surveillance authority by 2 August 2025. Belgium missed that deadline and has still not formally appointed one. While the Belgian Institute for Postal Services and Telecommunications (BIPT) was identified in the January 2025 Government Declaration and the 2025-2029 Federal Government Agreement as the intended main authority, formal legislation is still pending in parliament.

According to BIPT, the AI Act entered into force on 1 August 2024 and will be fully applicable from 2 August 2027, with obligations taking effect in stages. However, the institute’s formal designation as Belgium’s AI supervisor still requires legislative approval.

The FOD Economie, the Federal Public Service for Economy responsible for coordinating AI Act implementation, confirmed the legal vacuum. Etienne Mignolet of the FOD Economie stated: “The interpretation that there is currently a legal vacuum is correct. From 2 August, these rights can in the current state of affairs only be enforced through the courts.”

A Widespread Problem

The enforcement gap is not unique to Belgium. As Stijn De Roo, a CD&V senator who raised parliamentary questions on the issue, documented, Minister Clarinval confirmed that the FOD Economie is coordinating implementation and that “no new market surveillance authorities are being created for the AI Act.” The Netherlands also missed the designation deadline, and Mignolet noted the problem affects most EU member states.

Without a designated national authority, Belgian citizens cannot file AI Act complaints with a national body. Their only recourse is through the courts or the European Commission’s AI Act Service Desk. The European Commission acknowledged the issue just before the new phase began, stating that good enforcement “also depends on whether member states properly designate their national supervisory authorities and give them enough resources.”

The Digital Omnibus Provides Partial Relief

The recently adopted Digital Omnibus Regulation (EU) 2026/1744, in force since 27 July 2026, postpones high-risk AI system obligations from 2 August 2026 to 2 December 2027 (Annex III) and 2 August 2028 (Annex I). As ICT Rechtswijzer explains, the postponement “does not affect the transparency obligations” which remain on schedule. The Digital Omnibus also adds new prohibited practices from 2 December 2026, including bans on ‘nudify’ apps and AI systems generating child sexual abuse material.

For employers, the changes are significant. Partena Professional notes that while high-risk AI obligations for HR applications are postponed to December 2027, the transparency requirements still apply from today. The Central Council for Business also urged Belgian authorities to urgently establish proper AI oversight.

What to Watch For

Belgium’s federal structure adds another layer of complexity. As GLACIS notes, 21 specific bodies have been designated under Article 77 for fundamental-rights supervision - the highest count in any member state - but it remains unclear who oversees AI in education or employment, which are regional rather than federal competences.

The BIPT designation process could take several more months. The FOD Economie is working on a draft law that must pass through the full legislative process. Meanwhile, new prohibited practices take effect on 2 December 2026, and high-risk AI obligations follow in December 2027 and August 2028.

For now, Belgian citizens and businesses must navigate the new transparency rules without a dedicated national authority to turn to - relying instead on courts and the European Commission to enforce their rights under the EU’s landmark AI legislation.