From Sunday, companies operating in the European Union must clearly label content produced by artificial intelligence, including deepfakes and chatbot interactions. This marks the first major enforcement step of the bloc's comprehensive AI regulation, which aims to help citizens distinguish between authentic and synthetic material online.
The new obligations, part of the EU's AI Act, require that any image, audio, or video that has been generated or manipulated by AI be marked in a way that is easily detectable by users. This can be achieved through watermarks, metadata, or other technical markers. Chatbots must also inform users that they are interacting with an AI system, rather than a human.
Failure to comply can result in substantial fines, underscoring the EU's determination to enforce transparency in the digital sphere. The rules apply to professional and commercial uses of AI; individuals using AI for purely personal, non-commercial purposes are exempt.
Deepfakes and public-interest content
The EU has singled out deepfakes as a particular concern, given their potential to spread disinformation at scale. An EU official noted that generative AI enables the creation of false content "at unprecedented scale, tailored to specific audiences, and disseminated with remarkable speed." The official added that the rules are designed to "preserve citizens' ability to trust what they see, hear, and read."
In addition to deepfakes, the guidelines require that text published on matters of public interest be labelled if it was created using AI without human editorial oversight. This includes news articles or political commentary generated by algorithms without human review. Exemptions exist for artistic, creative, satirical, and fictional works, which are not subject to the labelling requirement.
The rules also cover emotion recognition and biometric categorisation tools, which must be disclosed when individuals are exposed to them. This is part of a broader effort to protect fundamental rights in the face of rapidly advancing AI technologies.
Existing AI systems have until 2 December 2026 to fully adapt to the new requirements, giving companies a transition period to implement the necessary technical measures.
Industry reaction and implementation challenges
The EU has faced criticism from some business groups who argue that the labelling rules add unnecessary regulatory burdens, especially as AI becomes more pervasive across the continent. However, supporters of the regulation contend that transparency is essential for maintaining public trust in digital content.
Ashley Casovan of the International Association of Privacy Professionals acknowledged the difficulties, saying, "We have heard that it is going to be very, very difficult to implement. But I think we often hear this with compliance requirements. And yet, the world turns and we figure these things out."
Major tech companies have already begun preparing for the new rules. TikTok has required creators to label AI-generated content for several years and reports that over three billion items have been labelled using its tools. Meta has introduced an "AI Info" label on Instagram and Facebook, while Google has signed the EU's code of conduct on AI transparency and is collaborating with firms like Nvidia, OpenAI, and Apple on digital tagging standards.
Despite these efforts, some industry figures warn of potential pitfalls. Karen Massin of Google cautioned about "regulatory complexity" that could prove counterproductive and "risks confusing the people these rules are meant to help."
The implementation of the AI Act's transparency provisions is being closely watched across the EU, as it sets a precedent for how the bloc intends to govern artificial intelligence. The success of these measures will depend on consistent enforcement and the development of reliable labelling technologies.
For more on the broader implications of the EU's AI regulation, see our earlier analysis of the transparency rules taking effect and the enforcement challenges ahead.
