The EU AI Act's transparency requirements are no longer on the horizon. Article 50 took effect August 2, 2026, and if you're running AI-generated creatives into European markets, non-compliance is an active risk — not a future one. The fine ceiling is €15 million or 3% of global annual turnover, whichever is higher.
Most teams running Meta ads into the EU are not ready. Here's what requires a label, how Meta's own system handles it, and what you need to do at volume.
In this post:
- What the EU AI Act actually requires from advertisers
- Which creative types trigger the disclosure obligation
- How Meta handles labeling — and where it doesn't cover you
- The three Ads Insights breakdowns that silently stopped working
- How to stay compliant when running AI-generated creatives at scale
What Article 50 Actually Requires
Article 50(4) of the EU AI Act requires deployers of AI systems that generate or manipulate "deep fake" content to disclose its artificial nature. The EU definition of a "deep fake" is broader than the common understanding: any AI-generated or AI-manipulated image, audio, or video content that resembles real persons, objects, places, or events and would appear authentic or truthful to an ordinary viewer.
For advertisers, this means:
| Creative Type | Disclosure Required? |
|---|---|
| AI-generated photorealistic product imagery | Yes |
| AI-generated realistic human models or faces | Yes |
| AI-cloned or synthetic voice in audio/video ads | Yes |
| AI-manipulated video (lip sync, face swap, deepfake) | Yes |
| AI-assisted enhancements (color grading, background cleanup, upscaling) | Generally no |
| AI-written ad copy (standard promotional text) | No |
The Commission Guidelines, finalized July 20, 2026, clarify that ordinary advertising copy isn't covered — the text obligation applies to AI-generated content addressing matters of public interest. Your product headlines are not in scope. Your AI-generated lifestyle images with realistic human models are.
How Meta Handles It — And Where It Falls Short
Meta signed the EU AI Act Code of Practice on Transparency in July 2026, and the platform does provide automatic "AI info" labels on content created using Meta's own generative AI tools within Ads Manager. If you're using Meta's built-in AI creative features — background generation, image expansion, text overlay enhancement — those labels get applied automatically.
That covers one part of your stack. It does not cover:
- Third-party AI-generated imagery (Midjourney, DALL·E, Firefly, Stable Diffusion, etc.) uploaded directly to Ads Manager
- AI-generated or AI-cloned voiceovers in video or audio ads (ElevenLabs, Eleven, Descript AI, etc.)
- AI-manipulated video produced outside Meta's tools
- AI-generated UGC-style content with synthetic human presenters
For all of these, the obligation sits with you — the advertiser. Meta's platform label does not apply retroactively or to content it didn't generate. If you built it with external AI tools and uploaded it, you are responsible for the disclosure.
For teams running AI-generated creatives at scale, this is where manual compliance breaks down fast. bulk handles the creative upload workflow directly — which makes it the right layer to enforce disclosure tagging before anything goes live. You set the policy once; it applies to every creative going into EU-targeted campaigns.
The Three Reporting Breakdowns That Stopped Working
Separately — but equally important for EU-focused accounts — three Ads Insights API breakdowns switched to opt-in status on August 6, 2026. If your reporting hasn't been audited since then, your dashboards may be silently returning zeroes.
The affected breakdowns:
- Device breakdown ()
impression_device - Hourly breakdown ()
hourly_stats - Frequency breakdown ()
frequency_value
The API returns HTTP 200 with empty data — no error, no warning, no notification. Accounts that never opted in stopped receiving data. If your ROAS-by-device or frequency analysis looks flat or suspiciously clean, this is why.
Enable opt-in through the new
insights/feature-settingsWhat Compliance Looks Like in Practice
For performance marketing teams and agencies running into the EU, a compliant workflow looks like:
EU AI Disclosure Compliance Checklist
- Audit your creative library: flag any assets generated with external AI tools (images, video, audio)
- Add explicit disclosure labels to AI-generated photorealistic imagery in ad descriptions or overlay text
- Ensure AI-cloned or synthetic voices in audio/video ads carry an audible or visible disclosure
- Do not rely on Meta's automatic "AI info" label for content you built outside Ads Manager
- Re-enable the three opt-in Insights breakdowns (device, hourly, frequency) per account
- Document your AI tool usage per campaign for compliance record-keeping
The documentation requirement matters. Regulators can audit. If you cannot show what content was AI-generated, which tools produced it, and how you disclosed it, the label on the ad itself may not be sufficient defense.
Agencies Running Multiple Accounts Face the Most Exposure
If you're an agency running 10, 20, or 50 client accounts into the EU, the compliance surface is large. Each account needs the Insights breakdowns re-enabled. Each client's creative library needs an audit. Each new AI-generated creative needs a disclosure decision before it goes live.
The teams doing this manually — creative-by-creative, account-by-account — are going to miss things. The teams that've built execution workflows where disclosure is part of the upload process won't. This is the same pattern that played out with GDPR consent requirements: the agencies that operationalized compliance early spent less time on remediation than those who treated it as a checklist.
The EU AI Act compliance guide from Billo covers the scope in detail. The cinerads breakdown of platform-by-platform disclosure rules is worth reviewing if you're also running TikTok or YouTube into the EU.
The standard for Meta ads running into Europe right now: if AI made it look real, you disclose it. No exceptions, and the August 2, 2026 deadline has already passed.
bulk handles the Meta ad execution layer — creative upload, campaign configuration, and A/B testing — for performance marketing teams running at scale. See how it works →