EU Introduces Mandatory AI Labels for Chatbots and Deepfakes

Wednesday, 05 August 2026 at 12:00

Europa voert verplichte AI-labels in voor chatbots en deepfakes

Since early August 2026, new transparency duties from Article 50 of the AI Act apply across the European Union. Any organization that develops or uses AI must clearly tell users in specific situations when they’re interacting with AI, or when content has been generated or altered by AI. The European Commission also released detailed guidance on how to apply the rules in practice.

The requirements don’t just hit makers of large AI models. Companies, governments, schools, platforms and other organizations deploying AI can all fall under the new rules.

What’s mandatory from 2 August?

The AI Act distinguishes between providers of AI systems and organizations that use them.

Providers must design systems so users know when they are directly interacting with AI. For a chatbot, voice assistant or digital avatar, it must be clear that no human is responding—unless that’s obvious to every user.

Providers of generative AI must also ensure that generated or altered text, images, audio and video carry a machine-readable marker. According to the European Commission, this technical tagging lets platforms and other systems detect AI content automatically, without always showing a visible label.

Additional disclosure rules apply to organizations using AI. They must inform users whenever they are exposed to:

emotion recognition or biometric categorization;
realistic AI deepfakes in image, audio or video;
AI-generated texts on matters of public interest when there is no human oversight and editorial responsibility.

Not every AI text needs a visible label

The rules are often misread as forcing a visible label on all AI-generated content. That’s not correct.

There’s a key exception for texts on topics of public interest when a human reviews the content and a person or organization remains ultimately responsible for publication. A newsroom that uses AI as a tool, verifies facts, and retains editorial responsibility doesn’t have to flag AI use above every article by default.

There are also exceptions for artistic, satirical and fictional works. Disclosures shouldn’t unnecessarily disrupt the experience of the work. However, for realistic deepfakes that make people appear to say or do things that never happened, clear transparency remains the default.

The Commission stresses in its guidance that obligations depend on the organization’s role, the AI system used, and the deployment context. The guidance and the voluntary General-Purpose AI Code of Practice support interpretation, but the legal transparency duties themselves are fully binding.

What organizations need to set up now

Compliance starts with knowing exactly where AI is used across the organization. An inventory of all AI applications is the first step.

Think beyond chatbots or text generators: include IVR and call automation, image generators, biometric features, and software with embedded AI that may be running in the background.

For each use case, it must be clear:

who is the provider and who is the user of the AI system;
which personal or other data are processed;
when users must be informed in advance;
how AI content is technically tagged or visibly labeled;
who is responsible for reviewing AI outputs;
how incidents, errors and complaints are logged.

These duties align with the broader AI literacy requirement in force since 2 February 2025. Organizations must not only inform staff that AI is used, but also ensure they understand what data can be safely entered, how to verify AI outputs, and when human intervention is necessary.

Fines up to 15 million euros

Organizations that breach the transparency rules face major penalties. Under Article 99 of the AI Act, fines can reach 15 million euros or 3 percent of global annual turnover, whichever is higher. For SMEs and start-ups, the lowest of the two caps applies.

Maximum fines aren’t automatic. Supervisors must consider the gravity and duration of the violation, the number of people affected, the organization’s size, remediation measures taken, and the degree of cooperation during the investigation.

Since 2 August, the European AI Office and national regulators oversee these provisions. The European Commission has also opened a complaints process and a whistleblower channel to report suspected AI Act violations.

Why these rules matter

The new transparency rules aim to stop AI from passing as human or as authentic media when that could mislead users. That makes Article 50 one of the first parts of the AI Act that people will notice in daily life.

For organizations, AI governance now goes far beyond picking a tool. They need visibility into where AI is deployed, the risks involved, who is accountable for outcomes, and how users are informed—clearly and in time.

The European Commission says these transparency duties build trust in AI without stifling innovation. Still, many organizations will have to adapt internal processes, documentation and communications to meet the new EU rulebook.