- Since August 2, 2026, Article 50 of the EU AI Act requires transparency for chatbots, synthetic content, deepfakes and emotion recognition.
- The rules can reach non-EU companies when their AI systems or outputs are used in the EU.
- Chatbots must tell people they’re interacting with AI unless it’s already obvious.
- Fines can reach €15 million or 3% of worldwide annual turnover; a grace period to December 2, 2026 applies to some existing generative systems.
If your website chatbot, AI phone agent or marketing content reaches people in Europe, a new set of rules now applies. On August 2, 2026, the transparency obligations in Article 50 of the EU Artificial Intelligence Act became applicable. Many U.S. small businesses assume EU law isn’t their problem — but if you sell online, book international customers or run a bilingual site, it’s worth ten minutes to check.
This article is general information, not legal advice. Talk to a qualified attorney about your specific situation.
What took effect
According to law firm analyses published after the deadline, Article 50 covers four situations:
- AI that interacts with people — such as chatbots and voice agents — must make clear that the person is interacting with AI, unless that is already obvious.
- Synthetic content — providers of systems that generate audio, images, video or text must mark outputs in a machine-readable way so they can be detected as AI-generated, with limited exceptions for standard editing.
- Deepfakes and certain AI-generated public-interest text — deployers must disclose that the content was artificially generated or manipulated.
- Emotion recognition and biometric categorisation — people must be informed when these systems are used on them.
Does this apply to a U.S. business?
It can. The rules apply to providers and deployers that place AI systems on the EU market or whose AI outputs are used within the EU. A Florida online store with a chatbot that serves customers in Spain, or a U.S. agency producing AI-generated ads for a European audience, may fall within scope. A local service business with no EU customers is far less likely to be affected — but the practices below are good for trust everywhere.
Deadlines and penalties
Content generated before August 2, 2026 does not need to be labeled retroactively. For generative AI systems that were already on the market before that date, a transitional period runs until December 2, 2026 for the marking and detection requirements; systems placed on the market on or after August 2 must comply from the start. Non-compliance can lead to fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher.
A practical checklist for small businesses
1. Inventory your AI touchpoints
List every place AI talks to customers or creates content: website chat widgets, AI receptionists, email assistants, product descriptions, ad images and videos. Note which ones can reach EU users.
2. Add a clear AI disclosure to chatbots and voice agents
A first message like “Hi, I’m the virtual assistant for [Business] — an AI. I can answer questions or connect you with our team” is simple and effective. Do the same at the start of AI phone calls. (It also improves customer trust — see what owners say about AI receptionists.)
3. Label AI-generated images and video
Ask your AI tools whether they embed machine-readable markers, and keep a visible note where content could be mistaken for real footage or photos. Google now also shows whether ads were made with AI.
4. Always offer a human path
Make it easy to reach a person — a button, an email or a phone option. It reduces frustration and shows the AI is a helper, not a wall.
5. Document what you use
Keep a one-page record of each AI tool: vendor, purpose, what data it sees and who owns it internally. If a customer, partner or regulator asks, you’ll have answers.
The bottom line
Transparency is quickly becoming the default expectation for AI in business. Disclosing clearly, labeling synthetic content and keeping humans reachable costs very little — and it protects both your compliance and your reputation. When we build chatbots and AI integrations for clients, these safeguards are part of the setup.