The EU AI Act: New Transparency Rules

On August 2, 2026, the transparency obligations in Article 50 of the EU AI Act started to apply. The core rule is simple: people have to know when they’re interacting with artificial intelligence, or when a piece of content was generated or manipulated with it. For a business, the first step isn’t to label everything, but to identify which systems you use and which obligation applies in each case.

What changed with the EU AI Act

The European Commission confirmed that Article 50 applies as of August 2, 2026. It distinguishes between providers —those who develop a system, or have one developed, and offer it under their own name— and deployers, meaning organizations that use AI for professional purposes.

The scope can include providers located outside Europe when their system’s output is used in the EU. A Latin American company with users or campaigns in Europe shouldn’t assume its location leaves it out: it should review its own case and, where appropriate, get legal guidance.

When you have to disclose or label

The official guidelines explain four main situations:

• Direct interaction with AI: a system designed to converse with people has to disclose that it’s AI from the start, unless that’s obvious.
• Synthetic content: providers of generative systems have to embed machine-readable marks so their output can be detected.
• Emotion recognition or biometric categorization: anyone using these systems has to inform the people exposed to them.
• Deepfakes and texts on matters of public interest: whoever publishes them has to clearly disclose their artificial origin in the cases covered.

Not every AI-assisted text needs a visible label. The Commission indicates that a text on matters of public interest can be exempt when it went through substantive human review or editorial control and someone is accountable for publishing it. A quick spell check doesn’t count as that review.

A practical checklist for your business

Before adding generic notices, document the actual flow:

1. List the chatbots, content generators and biometric tools you use.
2. Define whether your company acts as a provider, a deployer or both in each case.
3. Note where their output shows up: website, WhatsApp, social media, ads or support.
4. Check that every direct interaction clearly identifies the AI from the first contact.
5. Set up human review for sensitive content or content of public interest, and record who approves it.
6. Confirm with each provider what technical marking they embed.
7. Review territorial scope and exceptions with proper advice.

This inventory also improves the experience. A well-organized website lets you place notices, policies and contact channels where the user needs them, without turning every message into a legal block that’s hard to read.

Example: a chatbot and a social campaign

Imagine a company uses a chatbot to answer questions and a generative tool to create a campaign. In the chat, it has to show from the first message that the person is talking to AI when that isn’t obvious, and offer a clear path to human help.

For the campaign, the team has to distinguish between normal editing, synthetic content and possible deepfakes. Then it reviews whether a visible label, technical marking or documented editorial control applies. On a platform with a configurable agent and human control, like Nexlab Business, the reply and escalation rules can stay visible to the team; the legal assessment is still the business’s responsibility.

Frequently asked questions about AI transparency

Do I have to disclose it if I use an AI chatbot?

When the system interacts directly with people and it isn’t obvious that it’s AI, it has to disclose that from the start.

Does all AI-generated content need a visible label?

No. It depends on the content, the company’s role, the context and the exceptions. Machine-readable marking also doesn’t always replace a visible disclosure.

Can the AI Act affect a company outside Europe?

Yes, in certain scenarios. The Commission clarifies that outside providers can be covered when the system’s output is used in the EU.

Is there a grace period?

There’s a limited period until December 2, 2026 for certain systems placed on the market before August 2, and only with respect to the marking and detection of AI-generated content.

Transparency that builds trust

Complying with the EU AI Act starts with understanding the flow, not with copying a notice. Identify your systems, document your human oversight and communicate clearly when AI is involved. To organize support, rules and escalation, take a look at Nexlab Business and confirm the legal scope with a specialist.