Government & lawMixed

EU makes AI-generated content carry mandatory labels from August

New AI Act rules require visible labels and watermarks on realistic AI images, audio, video and text — with fines up to €15m for companies that ignore them.

By the Nuaico AI Desk·4 min read·August 1, 2026·Written and auto-published by AI — every source linked below

A European government office desk with a laptop screen showing a subtle black-and-white AI-content label icon overlaid on a video still, soft daylight from a window, documentary style, no readable text or logos, shallow depth of field.

What happened: Starting 2 August, new AI systems sold in the EU must label realistic AI-generated images, audio, video and text so people know they aren't authentic. Existing systems get four extra months. The rule comes from the EU's AI Act and targets deepfakes that have already been used to fake statements from politicians like Michal \u0160ime\u010dka, Emmanuel Macron and Keir Starmer.

Why it matters: Realistic fake audio and video have been used to smear politicians and could mislead voters, which lawmakers say threatens both individuals and democratic trust. EU officials frame the labels as protecting both consumers and the integrity of public debate, not just flagging novelty content.

How it works, plainly: Content designed to look real must carry a visible label plus an invisible digital watermark showing its AI origin. Text on public-interest topics needs labelling too, unless a human editor reviewed it. Personal content, satire, and clearly artistic or fictional work are exempt. The EU offers a standard black-and-white symbol, but companies can design their own.

The rollout: Meta, Google and over 180 other organisations have signed a voluntary code of practice to help meet the rules; signing is treated by regulators as good-faith compliance. Google's SynthID has already watermarked over 100bn images. Tech industry groups warn the labelling definition is broader than the original law intended, risking label fatigue similar to cookie banners.

The whole pictureEvery story cuts both ways. Here's this one.

The upside

Gives ordinary users a visible, consistent way to spot manipulated political content before it spreads unchecked.
Pushes major platforms to expand and standardize watermarking tools that already work at massive scale, like SynthID.
Creates real financial consequences (up to €15m or 3% of turnover) for companies that ignore transparency obligations.

The downside

Industry warns overly broad labelling could flood users with notices until people stop paying attention, much like cookie banners.
Enforcement leans heavily on voluntary codes of practice, so real compliance strength across all companies is untested.
Sectors like advertising, film, and publishing that quietly use AI at scale may face confusing new burdens with little public guidance.

Our read:a sensible transparency rule whose real test will be whether labels stay meaningful or just become background noise.

The ripple effect

Media — publishers and film/advertising must now flag AI use they previously kept quiet Tech & AI — platforms like Meta, Google and TikTok must scale watermarking and labelling systems Money — non-compliant companies risk fines up to 3% of global turnover

How this story was madeThis story was researched, written, illustrated and published by Nuaico's automated AI pipeline, with no human review before publication. Every source it drew from is linked below. Spotted an error? Email hello@nuaico.com and we'll fix it fast.

Sources

→ AI labels to be compulsory on authentic-looking content under EU rules (The Guardian)