A photo shows a famous person doing something shocking. Is it real, or was it made by AI? In the European Union, new rules aim to make the answer easier to find.
On August 2, 2026, transparency rules under the EU’s AI Act started to apply. Companies that provide generative AI systems must add special machine-readable marks to AI-created or AI-changed text, images, audio, and video. These marks may be hidden in the digital file, but online services and computer tools can detect them. Simple editing tools are generally excluded if they do not greatly change the original content or its meaning. (digital-strategy.ec.europa.eu)
Some content also needs a clear label that people can see. This includes deepfakes, such as a realistic fake video of a politician, and AI-written text about topics of public interest when no human editor has checked it. Chatbots must also tell users that they are talking to an AI system. The EU has created optional icons that organizations can use for these labels. (digital-strategy.ec.europa.eu)
The main purpose is to reduce deception and false information. The EU hopes that people will be able to make better decisions when they know how content was produced. Companies that break the rules may face large fines. In serious cases, a company can be fined up to €15 million or 3% of its total yearly worldwide sales. (digital-strategy.ec.europa.eu)
However, a label does not automatically make content safe or truthful. People still need to check the source, compare information, and think carefully before sharing. AI labels are an important new tool, but responsible online behavior remains necessary.










