EU Rules Set 3% Fines for Unlabelled AI Deep Fakes in Ads, Covering Non-EU Advertisers
TLDR
EU rules treat synthetic influencers and AI product shots that misstate an item as deep fakes requiring labels, with fines of up to 3% for non-EU advertisers, and older AI tools must add watermarks by December 2.
What this policy change means for your accounts
- Status
- Policy change
- Platform
- Not specified in the source
- Affects
- AI-generated ads, Synthetic influencers, AI product images
- Who
- Advertisers running AI-generated ads in the EU, including non-EU advertisers
- Effective
- December 2
- Action needed
- Label AI deep fake ads and make sure older AI tools add watermarks by December 2.
PPC Land reports that synthetic influencers and AI product images that misstate an item count as deep fakes under the rules, while ordinary ad copy is exempt. Non-EU advertisers face fines of 3% for unlabelled AI deep fake ads, and older AI tools must add watermarks by December 2.
How we got here: earlier coverage
- Nov 12Google Ads Will Suspend Advertisers Using AI Likenesses of Public Figures Without Warning
Google Ads will suspend advertisers on detection of AI likenesses of singers, athletes or YouTubers, with no warning, and AI labels will not exempt them.
- Nov 11EU Digital Fairness Act would bar ad profiling of minors and allow fines up to 6% of turnover
The EU's proposed Digital Fairness Act would stop ad tech firms, brands and agencies from profiling minors for ads on any website, with fines of up to 6% of turnover, but it cannot take effect until Parliament and the Council agree on it.
- Oct 28Meta Turns On Device-Side Conversion Counting by Default on November 12
Meta will switch on device-side conversion counting by default on November 12, and a scan of Meta ad files found only 2 of 3,379 AI-flagged files disclosed AI use.
- Oct 27Resemble AI Study Finds Nearly No Flagged Meta Ad Files Disclosed AI Use
A Resemble AI analysis of 509 brands found AI-generated media in one of every 11 Facebook and Instagram ad files, but only 2 of 3,379 flagged files disclosed AI use.
