The disclosure gap: what Article 50 actually means for your creative practice, and what to do before August 2

This article is for informational purposes only. It is not legal advice. If you have specific compliance questions, consult a qualified legal professional in your jurisdiction.

More than half of creative professionals using AI in client work haven’t told their clients about it. According to Envato’s “Beyond Adoption: The State of AI in Creative Work 2026” survey (n=1,780 creative professionals, November 2025), 58% are using AI in professional projects without disclosing that use. August 2, 2026, when the EU AI Act’s Article 50 transparency obligations take effect, is being treated by many as the compliance deadline that will fix this. It probably won’t.

What August 2 actually does is shift the professional norm. Client expectations around disclosure are already moving, and creatives who haven’t had this conversation are about to find it coming to them. The honest question isn’t whether you comply with a regulation. It’s whether your practice is ready for a conversation most clients are starting to ask.

What Article 50 actually requires of you

Most creatives assume the EU AI Act places sweeping disclosure obligations on them. The law is narrower than that, and understanding the distinction matters.

The Act draws a clear line between providers (the companies building AI tools: Adobe, Midjourney, OpenAI) and deployers (you, using those tools professionally). The technical marking obligation (embedding machine-readable watermarks and C2PA metadata) falls almost entirely on the provider. In most cases, when you use a third-party AI tool, that obligation is not yours to carry.

Your obligations as a deployer under Article 50 are more specific. They apply when:

• Your output is a deepfake: an AI-generated or manipulated image, audio, or video that falsely appears authentic. This means realistic portrayals of real people, places, or events that did not happen. Abstract AI illustration, AI-assisted logo design, or stylized generative imagery does not meet this definition.

• Your output is AI-generated text published to inform the public on matters of public interest: a narrow category that covers news-style content, not marketing copy or standard client deliverables.

An artistic or creative exception exists, but it has limits. Content that is “evidently artistic, creative, satirical, or fictional” still requires disclosure of AI origin, though you can do it in a way that does not disrupt how the work is experienced. Content that serves “primarily a commercial or informative purpose” loses that protection. AI-generated advertising imagery depicting real people or places cannot shelter behind the creative exception.

A note on sourcing: this analysis is based on the European Commission’s draft guidelines published May 8, 2026. The consultation on those guidelines closed June 3, 2026; final guidelines were pending publication at the time of writing. Check the EU AI Act Service Desk for updates before August 2.

For a broader overview of how Article 50 fits the Act’s wider framework, see the EU AI Act guide for creators on the TBM blog. For the full 7-step compliance checklist with tool-specific instructions, see: Your EU AI Act compliance checklist for creative teams

Why 58% aren’t disclosing

The Envato research asked non-disclosing creatives why. The most common answer, from 48% of that group: they don’t see why they need to disclose every tool they use.

That is not ignorance. It is a position, and it’s worth taking seriously before dismissing it.

Most creative work does not fall clearly within Article 50’s scope. A designer using Firefly to generate a background texture for a brand deliverable is not making a deepfake. A copywriter using ChatGPT to draft a press release and substantially rewriting it is probably protected by the editorial exemption. The instinct to say “this doesn’t really apply to me” is often legally reasonable.

The problem is professional, not legal. Clients are asking more frequently. Some clients (15% in the Envato survey, rising to 18–19% in the US and UK) explicitly want human-created, non-AI output, and they want to know before they hire you, not after delivery. When a client finds out their brief was processed through an AI tool without any mention of it, the issue is rarely the tool itself. It’s the omission.

A separate 2026 academic study reviewing how creators discuss AI use on professional platforms found a consistent pattern: creatives acknowledge using AI internally but default to non-disclosure because they don’t know whether clients will object, and they’d rather not start a conversation that might cost them the project (CHI 2026, arXiv:2603.07459). This is the “ask forgiveness not permission” dynamic. It predates the law entirely.

What changes on August 2, honestly

Enforcement infrastructure across the EU is still being built. As of May 2025, only three member states (Lithuania, Luxembourg, and Malta) had fully designated the national competent authorities required to enforce the Act. Fourteen states had designated none. Most regulatory analysts describe the realistic 2026 picture as a guidance and warning phase, with enforcement focused on large platforms and systematic violations. For an individual freelance creative, the risk of a fine in 2026 is low.

That framing is honest. It is not an argument for doing nothing.

What does change on August 2: client conversations. Procurement processes at larger organizations are already updating supplier questionnaires to include AI use disclosure. Request-for-proposal documents are beginning to include explicit AI policy requirements. Creatives who have sorted their disclosure practice (who have a clear answer when asked) are finding that transparency functions as a differentiator, not a liability.

The August 2 date also carries a signalling effect independent of enforcement. For clients who have been uncomfortable with AI but unsure how to raise it, a live legal obligation gives them a clear reason to ask. The creatives who haven’t thought about this are about to find the conversation comes to them.

Four things to do this month

This is not the full compliance checklist. If you want the step-by-step guide with tool-specific instructions, it’s here: Your EU AI Act compliance checklist for creative teams.

These four things take under two hours and prepare you for the conversation:

Know which of your outputs are actually in scope. Go through your current projects. If none involves a realistic AI-generated portrayal of a real person or place, or AI-generated text published as public interest journalism, Article 50’s disclosure obligation probably does not apply to your day-to-day work. That is a clarifying exercise, not a compliance shortcut.

Check your tools’ C2PA status. Adobe Firefly and DALL-E 3 embed C2PA metadata automatically. Midjourney does not. Knowing which of your tools do the technical marking for you tells you where your manual disclosure obligations sit.

Prepare one disclosure sentence per channel. “This image was generated using AI.” “This video contains AI-generated visuals.” Brief, visible, present at the point of delivery or publication. You don’t need a paragraph. You need a sentence, ready to go.

Update your client conversation before delivery. If you use AI tools in client work, say so before the work lands, and before August 2. Not as a confession. As a matter of professional practice. Most clients will not object. A few will ask questions. Almost none will respond worse to honesty than to a surprise.

The gap is older than the law

The disclosure gap is not a compliance problem. It is the industry’s oldest problem wearing new clothes: the gap between what professionals do in practice and what they say they do. August 2 will not close that gap. What closes it is deciding your practice has a position, and being willing to defend it.