EU to Require Labels on Realistic AI Images From Sunday

Starting Sunday, the EU will require tech companies to label AI-generated content that impersonates people or real events under the AI Act, with fines for non-compliance. The rules target deepfakes designed to deceive, exempting artistic, satirical, or private use, and place responsibility on platforms for content from outside the EU.

EU to Require Labels on Realistic AI Images From Sunday The EU’s AI Act https://petapixel.com/2024/03/13/eus-landmark-ai-act-has-been-approved-us-far-behind/ is being introduced in stages, and this coming Sunday any AI-generated content impersonating people or real events will be required to have a label. It means that tech companies, including social media giants, must now clearly mark a deepfake or else face heavy fines. Le Monde reports https://www.lemonde.fr/en/international/article/2026/07/28/eu-tells-firms-to-label-ai-generated-content-from-sunday 6755910 4.html that this can be done by including watermarks or labels — any method is acceptable so long as it allows for the easy detection of synthetic content. The EU has particular concerns about deepfakes that imitate reality and are designed to fool people. “Generative AI enables disinformation to be created at unprecedented scale, tailored to specific audiences, and disseminated with remarkable speed,” says an EU official, per Le Monde . The official adds that the new rules are aimed at helping citizens maintain trust in what they see, hear, and read. The rules only relate to content that is made for professional purposes, and it extends to informational text designed for public consumption that was AI-generated without any human oversight. AI content that is made for artistic, satirical, or fictional reasons falls outside of this legislation’s scope. People who create AI content for private use also won’t be affected. Google has signed the EU AI Act code of conduct on AI transparency, while Meta has signaled its intention to sign. Google says that while it supports transparency, it is “concerned that adding more regulatory complexity — as technical solutions are still evolving — could contradict Europe’s goals for competitiveness and simplification.” Meta, meanwhile, says that “We must avoid a situation where online content is subject to a growing array of different labels and disclosures that end up overwhelming people, while at the same time adding additional regulatory complexity for providers.” Meta introduced an “AI Info” label https://petapixel.com/2024/07/01/instagram-changes-made-with-ai-label-to-ai-info-after-widespread-backlash/ in 2024, while TikTok has required content creators to label AI imagery for several years. So it is likely that major platforms believe they already have the tools in place for the EU’s impending new rules. While these new rules affect people in the EU, someone outside of the EU could make a deepfake that is quickly viewed by European citizens. In that case, the onus falls on the platform. If their AI detection tools fail, what then? The companies also say that AI is a moving target. What works now might not work in a few months’ time. Currently, big tech relies on tools like the C2PA protocol https://petapixel.com/2022/01/26/the-first-standard-to-assure-a-photos-authenticity-has-been-created/ and the invisible SynthID watermark. All of these systems have flaws. Image creditsHeader photo licensed via Depositphotos https://depositphotos.com/ .