Every AI Ad, Chatbot and Playlist in Europe Now Needs a Label
The EU's AI Act transparency rules kick in on 2 August. From Spotify recommendations to complaint hotlines, businesses must tell people when AI is involved or face fines of up to €15 million.
Key points
- The EU's Artificial Intelligence Act transparency rules take effect on 2 August 2025, requiring businesses to disclose when AI is involved in content or services.
- Companies that fail to comply face fines of up to €15 million (about $17 million) or 3 percent of global annual revenue, whichever is higher.
- Affected businesses include obvious AI firms and companies like Spotify (music recommendations) or Adobe (photo editing tools).
- A transitional period runs until December 2025 for labelling AI-generated audio, images and video in a machine-readable format.
- Enforcement across the EU's 27 member countries may be slow and uneven in the early months, lawyers warn.
If you live in the EU and a chatbot helps you file a complaint this week, it now has to say so upfront. Starting 2 August 2025, businesses must tell people whenever AI is involved in what they see or interact with. That's the short version of the new transparency chapter of the EU's Artificial Intelligence Act.
What exactly has to be labelled?
Quite a lot. Ads built using deepfakes (fake images or videos generated by AI) must carry a label. Customer-service chatbots must identify themselves as AI. Call centres using AI to monitor a caller's emotional tone must say so at the very start of the call.
The rules reach into business-to-business dealings too. Scheduling by AI, automatic email handling, AI-assisted contract negotiation: all of it needs to be declared. Technology lawyer Thibau Duquin of Stibbe told Wired: "When you start looking, it's everywhere, and all this content will have to be labeled from now on."
Spotify's AI-driven listening recommendations count. So does the generative fill button in Adobe Photoshop. The European Commission's AI Office, which we first covered on 2 August 2026, will watch over major model developers including OpenAI and Google DeepMind.
Will it feel like more annoying pop-ups?
Possibly, and critics are already worried. Boniface de Champris, AI policy lead at the Computer & Communications Industry Association, flagged the risk of "banner blindness" back in December. His concern: if a spell-checked email and a filtered selfie both carry warnings, the warnings stop meaning anything.
The comparison with the EU's GDPR, the General Data Protection Regulation that governs personal data, is hard to ignore. That law came into force in 2018 and gave consumers the right to know how companies used their personal data. The result was the cookie-consent banner that now greets you on almost every website, and the "cookie fatigue" that followed.
| Rule | Deadline | Example |
|---|---|---|
| Chatbots must identify as AI | 2 August 2025 | Customer service bots |
| AI-generated ads must be labelled | 2 August 2025 | Deepfake marketing images |
| Emotion-monitoring AI on calls | 2 August 2025 | Call centre frustration detection |
| Machine-readable labels for synthetic media | December 2025 | AI-generated audio or video |
What happens next?
Enforcement will be patchy to begin with. Each of the EU's 27 member countries is at a different stage of setting up its own oversight body. Rosie Nance, a data and AI regulatory lawyer at Norton Rose Fulbright, told Wired that consistent policing will take time, and warned that "we're not necessarily going to see consistent enforcement straight away."
For businesses, the bigger question is strategic. Do they label their AI use and carry on, or do some quietly go back to doing things without AI to avoid the transparency? Duquin told Wired he sees it as a genuine fork in the road. For ordinary people, the practical upshot is this: if labels are missing from a service you suspect uses AI, that absence itself becomes meaningful. Our earlier story on what the labelling rules mean for AI-generated images and audio is worth reading alongside this one.
Common questions
Does this affect apps I already use?
Yes. Any app or website that serves EU users and uses AI in its recommendations or customer service is covered, including global platforms like Spotify and tools like Adobe Photoshop.
What if a company does not comply?
Fines can reach €15 million or 3 percent of a company's global annual revenue, whichever figure is higher. Enforcement will initially vary by country, but penalties are on the table from 2 August 2025.
Does this apply to personal use, like a filter on my own photo?
The rules target providers and businesses, not individual consumers editing a personal photo. If a company publishes that AI-edited image as a commercial ad, however, that company must label it.
