EU AI Act Transparency Rules Take Effect August 2: Two-Tier Labeling, Fines Up to €15M

Starting August 2, 2026, the EU AI Act requires AI-generated content to be labeled with machine-readable marks and, for deepfakes and public-interest texts, visible labels. Non-compliance can cost up to €15 million or 3% of global turnover.

This article was drafted with AI assistance from multiple sources and was reviewed and approved by a human editor before publication.

The European Union's AI Act transparency obligations will come into force on Sunday, August 2, 2026, mandating that AI-generated and AI-modified content be clearly labeled. The rules apply to both AI system providers—such as OpenAI, Google, and Meta—and professional users who deploy the technology.

The regulation introduces a two-tier labeling approach. First, providers must embed machine-readable technical marks—such as metadata, digital watermarks, or other provenance traces—in AI-generated images, videos, audio, and text. These marks must be effective, interoperable, reliable, and as resilient as technically possible. Second, professional deployers must add a visible or audible label to deepfakes: content that resembles real people, objects, places, or events and could appear authentic. The same visible labeling is required for AI-generated text that informs the public on matters of public interest, including politics, health, safety, justice, consumer issues, science, and culture.

Exemptions exist. Text labeling is not required if a human reviewed the content before publication, editorial control is exercised, and a legal entity takes responsibility. Purely fictional, artistic, satirical, or creative works are exempt from visible labels, though technical marking still applies. Private, non-professional users are not subject to visible labeling obligations.

There is no single uniform label; the European Commission has published examples, but each provider may use its own label. The Commission also issued guidelines to support implementation, as noted by Henna Virkkunen, Executive Vice-President for technological sovereignty: "With the guidelines published today, the Commission supports the smooth and effective application of the AI Act to make AI systems that interact with people more transparent and trustworthy – such as chatbots, AI agents, and AI content."

Penalties for non-compliance can reach up to €15 million or 3% of global annual turnover. Enforcement will be carried out by national supervisory authorities, the EU AI Office, or the European Data Protection Supervisor.

A grace period applies: AI systems already on the market before August 2, 2026, have until December 2, 2026, to comply. Content generated before the effective date is not subject to the labeling obligations.