The EU AI Act's Article 50 Transparency Rules Go Live August 2. Here's What to Fix First

If your product has a chatbot, generates images or video, or produces AI-written text that reaches users in the European Union, you have about two and a half weeks before Article 50 of the EU AI Act stops being a compliance roadmap item and starts being an enforceable law. The obligations go live August 2, 2026, across all 27 member states, and the penalties (up to €15 million or 3% of global annual turnover) apply regardless of where your company is headquartered.

This isn’t the AI Act’s high-risk-system regime, which has its own separate and much heavier compliance burden, and which got real breathing room this year when the Digital Omnibus pushed standalone high-risk obligations from August 2026 to December 2027. Article 50 is a different, narrower track that the omnibus delay did not touch. It’s about disclosure, telling people when they’re talking to a machine or looking at synthetic content, and it’s still on its original schedule. If you read the high-risk delay news and assumed the whole Act got pushed back, that’s the mistake to fix before August 2. Article 50 is narrower and more mechanical than the high-risk regime. That narrowness is exactly why it’s easy to underestimate. It reads like a UX tweak. Missing it is a fine.

The four things Article 50 actually covers

Chatbots and conversational AI. Any system designed to interact directly with a person in a way that could plausibly be mistaken for a human must disclose that fact at the first point of contact. Not in a footer link. Not in terms of service. The disclosure needs to be visible or audible before or as the interaction begins.

Emotion recognition and biometric categorisation. Systems that infer emotional state or biometric categories (whether someone appears distressed, their inferred age range, and similar signals) must tell the people being analyzed that this is happening.

Deepfakes and synthetic media. AI-generated or manipulated audio, image, or video content that resembles real people, places, or events needs a disclosure that a reasonable person would actually notice, whether that’s a persistent visual label, an opening disclaimer for video, or an audible cue for audio.

AI-generated text on matters of public interest. Text generated or manipulated by AI that informs the public on matters of public interest needs disclosure that it’s AI-generated, unless a human has reviewed and taken editorial responsibility for the content.

Why “we have a privacy policy” isn’t compliance

The most common mistake product teams are likely to make here is treating this like a documentation requirement instead of a UX requirement. A chatbot disclosure buried in an FAQ page technically exists, but it doesn’t meet “clearly informed… at the first point of contact.” The guidance from the European Commission points toward layered disclosure: a plain-language notice, plus a persistent visual indicator for ongoing interactions, plus audio cues where relevant.

For a support chatbot, that likely means a visible “You’re chatting with an AI assistant” banner at the start of the session, not just a one-time modal a user can dismiss and never see again. For AI-generated marketing images that resemble real people, it means a label that survives being screenshotted and shared, not metadata that only shows up if someone inspects the file.

The July 22 signatory deadline is separate and voluntary

There’s a second date worth knowing about that’s easy to confuse with the enforcement date: July 22, 2026, a deadline to sign the EU’s voluntary Code of Practice for AI content transparency. Signing gives companies a safe-harbor compliance pathway with clearer, pre-approved implementation guidance. It’s not mandatory, and not signing doesn’t exempt anyone from the underlying legal obligation that starts August 2. But for teams still deciding how to implement disclosure, the Code of Practice is worth reading before you build your own approach from scratch, since it’s the closest thing to an official implementation checklist that exists right now.

A practical checklist before August 2

Area	Question to ask	Action if the answer is “yes, and we haven’t handled it”
Chatbot / assistant	Does it interact with EU users in a way that could be mistaken for a human?	Add a clear, persistent AI disclosure at session start
Generated images/video	Could outputs be mistaken for real people, places, or events?	Add a visible label that survives sharing/screenshotting
Generated audio	Could outputs be mistaken for a real person’s voice?	Add an audible disclosure at the start of playback
Generated text/content	Does it inform the public on matters of public interest without human editorial review?	Disclose AI generation, or add a human review step that takes editorial responsibility
Emotion/biometric inference	Does your system infer emotional state or biometric categories from users?	Disclose this to the people being analyzed

Most teams building on top of the major LLM APIs already have the underlying capability to add this disclosure; the gap is almost always product decision-making, not engineering effort. If you’re scoping this work and want to make sure the disclosure implementation actually satisfies “clearly informed at first point of contact” rather than just checking a box, that’s a conversation worth having with legal counsel alongside your product team, not engineering alone.

What this means if you’re shipping AI features fast

Teams that have been moving quickly on AI features, which is most teams right now, are the ones most likely to have shipped a chatbot or content-generation feature without thinking through EU-specific disclosure requirements, because the requirement didn’t exist when the feature launched. If you’re auditing your product surface against this deadline and need engineering capacity to implement disclosure UX correctly across every affected touchpoint before August 2, that’s exactly the kind of focused, time-boxed work our team takes on. Waiting until after the enforcement date to find out you missed a surface is the expensive way to learn about Article 50.