The European Commission published its final implementation guidelines for Article 50 of the EU AI Act today, July 20, 2026. The August 2 compliance deadline is now 13 days away.

This is not another consultation document. These are the binding guidelines that tell AI providers and deployers exactly what the law requires from them — right before it becomes enforceable.

European Commission Executive Vice-President for Technological Sovereignty, Henna Virkkunen, put it plainly in the announcement: “Europeans have a right to know whether what they see, hear or read has been made or altered by AI, especially when such content can shape public debate.”

What Article 50 Actually Requires

The guidelines draw a sharp line between two categories of business, and the obligations are different depending on where you sit.

Providers of AI systems — companies that build AI models, APIs, or AI-powered products — must ensure their systems are technically capable of disclosing that users are interacting with AI. For interactive systems like chatbots, voice agents, and AI assistants, the disclosure must be automatic and clear. For generative AI systems that produce content, providers must embed machine-readable signals (watermarks, metadata, content credentials) so the AI origin can be detected downstream.

Deployers of AI systems — businesses that use AI tools to interact with customers or produce content for public use — must ensure visible labelling when AI-generated content touches areas like public information, employment decisions, financial guidance, or healthcare. If your AI agent talks to customers, handles support tickets, drafts client-facing reports, or produces content on matters of public interest, deployer obligations apply to you.

The guidelines also clarify a grace period: for AI systems already on the market before August 2, the content marking and detection obligations don’t apply until December 2, 2026. But the disclosure requirements for interactive AI — telling users they’re talking to an AI — apply from August 2 regardless.

Why This Matters More Than You Think

Most businesses with EU customers are running AI across more workflows than they realise. An AI agent that handles customer enquiries, a chatbot on your website that answers product questions, an AI-generated email sequence to a European prospect list, a voice assistant that takes bookings — all of these sit inside the scope of Article 50.

The guidelines confirm that “matters of public interest” under the AI Act include employment, financial guidance, health information, and legal communication. That covers a large percentage of what enterprise AI is actually doing today.

The penalty exposure under the EU AI Act for non-compliance is not trivial. Transparency violations can attract fines of up to €15 million or 3% of annual global turnover, whichever is higher. Regulators in Germany, France, and the Netherlands have already signalled active enforcement intent for 2026.

What Businesses Need to Do Before August 2

The guidelines are detailed, but the business checklist is straightforward.

Map your AI touchpoints. Every AI-powered interaction with EU-based users or customers is in scope. Start with your customer-facing AI — chatbots, voice agents, automated email responses, AI assistants embedded in products — and work outward to internal tools that produce content shared with EU stakeholders.

Audit your disclosure language. When a user interacts with your AI chatbot or voice agent, do they know it’s an AI? The disclosure must be clear, timely, and not buried in a privacy policy. If your AI agent introduces itself as “Emma from Support” without clarifying it’s an AI, you have a gap.

Check your AI vendor’s content credential support. If you use third-party AI to generate content — text, images, audio — does that provider support machine-readable signals (C2PA content credentials)? This matters for the December 2 marking deadline. If your AI tools can’t produce credentialed outputs, you need to plan for a switch.

Get your documentation in order. Regulators in enforcement proceedings look for evidence of assessment and intent to comply, not just end-state compliance. A documented audit of your AI interactions — even one that surfaces gaps you’re actively fixing — is far better than silence.

The Gap Between Knowing and Doing

The uncomfortable fact is that Article 50 has been in the text of the EU AI Act since 2024. The August 2 deadline has been known for over a year. Most businesses haven’t moved because the guidelines weren’t final.

Now they are. The businesses that have been waiting for this moment have 13 days to prove the gap was preparation, not procrastination.

The businesses that haven’t been tracking this at all are in a different position. Thirteen days is not enough time to build a comprehensive compliance programme from scratch. It is enough time to understand your exposure, document what you have, and put a credible remediation plan in writing.

What This Means for Business

AI regulation is not the obstacle it’s framed as. The disclosure requirements under Article 50 are asking businesses to do something that most customers already expect: be honest about when they’re talking to a machine.

Voice agents that disclose their AI nature upfront typically see no meaningful drop in customer satisfaction — and they avoid the trust collapse that happens when a customer discovers mid-conversation that “Sarah from Customer Support” was never a person.

The businesses that implement clear AI disclosure don’t just avoid regulatory risk. They build the kind of trust that converts into retention, especially in regulated industries where every customer interaction carries liability weight.

Enterprise DNA works with businesses to map their AI operations, identify compliance exposure, and design disclosure frameworks that meet regulatory requirements without degrading the customer experience. If you’re running AI in Europe and haven’t confirmed your August 2 position, book a discovery call this week — not next week.