5 Ways Your Chatbot Disclosure Fails Article 50 (Even If You Think You're Compliant)

Last updated: 25 June 2026

Guidance only, not legal advice. Based on EU AI Act Regulation (EU) 2024/1689, the Commission's draft Article 50 guidelines (May 2026), and the EU AI Office Code of Practice (June 2026).

You've done the right thing. You read about Article 50, you added some kind of AI disclosure to your chatbot, and you moved on. But the regulation is more specific than most business owners realise — and a disclosure that exists is not the same as a disclosure that complies.

Here are five ways businesses think they've handled it and haven't — and exactly what to fix.

1Your disclosure is in your privacy policy or terms of service

❌ The failure: "We use AI-powered chat tools to assist customers" buried three paragraphs into your privacy policy. Most businesses add this and call it done.

✓ The fix: Article 50(1) requires the disclosure to appear "at the latest at the time of the first interaction." The Commission's draft guidelines are explicit: a clause in your terms of service or privacy policy does not satisfy this requirement. The disclosure must be at the point of interaction — either in the chatbot's opening message, or in a persistent label visible before or at the start of the conversation.

The right placement is either "Hi! I'm an AI assistant" as the chatbot's first message, or a visible label near the chat widget that's present before the user types. Not a footer. Not a policy page. Not an FAQ.

2The text is there but it's too small, grey, or hidden

❌ The failure: A tiny "Powered by AI" label in 10px grey text under the chat widget that nobody reads. Or an "AI" badge so small it's invisible on mobile.

✓ The fix: Article 50 requires the disclosure to be "clear and distinguishable." The Commission's draft guidelines specifically cite accessibility requirements — the disclosure must meet applicable accessibility standards, meaning adequate contrast, readable font size, and visibility across devices.

The EU Commission guidelines explicitly note that "a very small snippet of text hidden in the footer of a website" or "a faint label on an image" would not qualify as clear and distinguishable. If your disclosure could be missed by a typical visitor on a mobile screen, it likely doesn't meet the standard.

3Your chatbot provider shows a label — you assume that covers you

❌ The failure: Tidio, Intercom, or Gorgias shows "Powered by AI" somewhere in their interface. You've checked their compliance page. You assume you're covered.

✓ The fix: Article 50 places the disclosure obligation on the deployer — you, the business running the website — not on the chatbot platform. Your provider may show a generic label, but you remain responsible for ensuring that disclosure is clear, distinguishable, and fires correctly on your specific site. More critically: you have no evidence log showing that the disclosure fired for your visitors on any specific date. That documented record is yours to maintain.

This is the most common misconception. Your vendor's compliance documentation protects your vendor. It doesn't protect you.

4You have no proof the disclosure actually fired

❌ The failure: You've set up the disclosure. It shows on your site today. But if a national authority opens a review next month and asks you to prove the disclosure was showing from 2 August 2026 onwards, you have nothing — no log, no timestamped record, no audit trail.

✓ The fix: The EU AI Act's enforcement mechanism under Article 99 asks authorities to consider documented good-faith compliance as a mitigating factor. That means having a record — not a promise, not a screenshot of settings, but a timestamped log showing the disclosure fired, when, and how many times.

This is the evidence gap that most businesses discover only when they're under review. A correctly worded disclosure with no proof it ran is legally equivalent to no disclosure. The record is the compliance.

This is exactly what Disclo Pro's evidence log handles automatically — server-side logging of every time your disclosure badge fires, with one-click export for any regulator request.

5You're assuming the "obvious from context" exception applies to you

❌ The failure: "My chatbot is clearly a bot — it has a robot icon and the name 'Helpbot'. Everyone knows it's not human." Relying on this to skip the disclosure.

✓ The fix: Article 50(1) includes a narrow exception for cases where the AI nature is "obvious from the circumstances and context of use." The EU Commission's draft guidelines use an "average consumer" standard — not a tech-savvy user, but an ordinary person. The guidelines give concrete examples: AI coding assistants available only to developers may qualify. Customer service chatbots on websites do not — even with a bot name or icon. Natural language chatbots that could be mistaken for a human at first glance are not "obvious."

The Commission explicitly warns against over-relying on this exception. The cost of a disclosure is one line of text. The cost of incorrectly relying on the exception — and being found in breach — is up to 3% of global annual turnover.

The quick compliance check

Run through these five questions about your current setup:

Does your AI disclosure appear at or before the very first chatbot message — not in a policy document?
Is it visually clear and distinguishable — adequate size, contrast, visible on mobile?
Do you have your own evidence of this — not just a screenshot of your vendor's settings, but a timestamped log of disclosure events?
Are you confident the "obvious from context" exception doesn't apply to your chatbot?
Is your disclosure in the right language for your EU visitors?

If you answered "no" or "not sure" to any of these, you have a compliance gap. The free Article 50 Validator at disclo.eu will tell you exactly what applies to your setup in under two minutes.