Key Developments

On June 10, 2026, the European Commission published its Code of Practice on marking and labelling AI-generated content—a voluntary guidance tool designed to help providers and deployers of generative AI systems meet transparency obligations under the EU AI Act. The release comes as the bloc moves toward a critical August 2026 milestone when transparency rules formally take effect across the EU.

The Code of Practice addresses a core requirement: marking AI-generated content and disclosing the artificial nature of images, audio (including deepfakes), and text. Rather than imposing prescriptive rules, the Commission opted for a guidance-based approach—signalling pragmatism around how companies operationalise transparency in practice.

This follows a broader push to streamline AI Act compliance. On June 3, the Commission formally adopted a legislative proposal for the Cloud and AI Development Act (CADA), a sovereignty-focused framework that will reshape how EU public institutions procure cloud and AI services. The dual initiatives—transparency guidance and procurement reform—reflect the EU’s attempt to balance regulatory rigour with practical market realities.

Industry Context

The August 2026 transparency deadline represents the first major enforcement wave under the AI Act since prohibitions on high-risk practices took effect in February 2025. For builders, this shifts the focus from policy abstraction to operational necessity: deepfakes, content labels, and disclosure mechanisms move from discussion points to compliance requirements.

The voluntary Code of Practice is intentionally non-binding. This matters. It signals the Commission recognises one-size-fits-all mandates won’t work across gaming, social media, enterprise software, and professional AI tools. Providers now have room to design disclosure approaches suited to their use cases—provided they can justify their choices under the Act’s transparency obligations.

Practical Implications

For Irish and European AI builders and deployers, the August deadline creates a hard checkpoint. Your compliance roadmap should now include:

  • Content classification: Map every AI output in your system (generated text, images, audio, synthetic media) and determine disclosure requirements.
  • User interface updates: Plan how you’ll surface transparency information without degrading user experience—the Code provides examples but leaves design decisions to you.
  • Documentation: Prepare technical records showing how your systems meet marking and labelling standards. Market surveillance authorities, including Ireland’s incoming AI Office of Ireland, will expect this.
  • Liability readiness: Understand that failures to disclose AI-generated content can trigger fines up to €35 million or 7% of annual worldwide turnover.

Open Questions

Key uncertainties remain. The Code is guidance, not law—its enforceability depends on how national authorities and the European AI Office interpret it. Ireland’s AI Office, expected to be operational by August 1, 2026, will shape how Irish regulators apply these standards. How different sectors (social media vs. enterprise) will implement disclosure varies widely, and consistent enforcement across member states is unclear.

The voluntary approach also raises a practical question: if adoption of the Code is uneven, will the Commission move toward binding standards? Monitor European Commission and AI Office guidance as August approaches for clarification.

Source: European Commission