Key Developments

The European Commission has published draft guidelines on the implementation of transparency obligations under Article 50 of the EU Artificial Intelligence Act, opening a targeted consultation period that closes on June 3, 2026. This marks the first comprehensive Commission instrument to provide interpretive guidance across the full scope of transparency requirements—coming less than three months before these obligations become binding on August 2, 2026.

The transparency rules include interactive AI disclosure requirements, restrictions on emotion recognition and biometric categorisation, and deepfake labelling obligations. The final Code of Practice on detecting and labelling artificially generated content is also expected in late June 2026.

Simultaneously, the provisional agreement reached on May 7, 2026, between the Council of the European Union, the European Parliament, and the European Commission on the Digital Omnibus amendments has streamlined certain compliance timelines. High-risk AI system obligations have been postponed from August 2, 2026 to December 2, 2027—a 16-month deferral—while product-embedded systems get a one-year extension to August 2, 2028. This pragmatic adjustment reflects recognition that standards development is still ongoing.

Industry Context

The EU AI Act represents the world’s first comprehensive AI regulation, setting a precedent that other jurisdictions are watching closely. With Ireland as an EU member and home to major tech companies, the stakes are particularly high for Irish and European AI developers and deployers.

The transparency rules are particularly significant because they apply directly to a broad category of AI systems already on the market. Unlike high-risk system requirements, transparency obligations don’t hinge on risk classification—they apply to generative AI systems placed on the EU market from August 2, 2026 onwards.

Practical Implications

For builders in Ireland and across the EU, the June timeline matters immediately:

  • June 2026: Final Code of Practice and transparent AI guidelines will be published. Organisations should review these carefully to understand practical compliance expectations.

  • August 2, 2026: Transparency rules become binding. Systems placed on the market from this date must comply immediately. Systems already on the market before this date face different timelines depending on which obligations apply.

  • August 2027: High-risk system obligations begin (under the revised timeline), but organisations should use 2026 to prepare, as standards are still being finalised by CEN and CENELEC.

Companies should audit which obligations apply to their systems now: Does your AI generate text, images, audio, or video that users might mistake for human-created content? If yes, you’ll need detection and labelling mechanisms ready by August 2, 2026.

The deferral of high-risk system rules provides breathing room but also creates urgency. Standards remain incomplete, but the Commission expects them to be published in the second half of 2026 or first half of 2027.

Open Questions

  • How will the “obviousness” exception for interactive AI systems (Article 50) be applied in practice? The guidelines provide examples, but real-world edge cases will emerge.

  • Will the final Code of Practice adequately address multimodal AI systems that generate multiple content types simultaneously?

  • How will national authorities coordinate enforcement across the EU, given that the AI Office has exclusive competence only for certain system categories while member states oversee others?

  • What happens to systems that are substantially modified after August 2, 2026—do they need to comply with rules in force at the time of modification or placement on the market?

Source: European Commission