\n<\/p>\n
Deepfakes and other AI-generated content must carry clear labels from Sunday, as the European Union\u2019s transparency rules under the AI Act come into force.<\/p>\n
From 2 August 2026<\/strong>, companies operating in the EU must ensure their AI systems \u2014 including chatbots \u2014 clearly disclose that users are interacting with artificial intelligence. Any image, video, audio or text created or significantly manipulated by AI must also be labelled so people can immediately tell whether the content is real or synthetic.<\/p>\n The goal is straightforward: help Europeans distinguish authentic material from AI-generated content at a time when generative AI can produce disinformation at unprecedented scale and speed.<\/p>\n The new obligations, set out in Article 50 of the AI Act, apply in stages and cover four main situations:<\/p>\n Firms that fail to comply face significant fines \u2014 up to \u20ac15 million or 3% of global annual turnover, whichever is higher.<\/p>\n Compliance can be achieved through a combination of technical and visible measures:<\/p>\n Existing generative AI systems already on the market before 2 August 2026 have a limited grace period until 2 December 2026<\/strong> to implement the machine-readable marking requirement. All other transparency duties apply from 2 August.<\/p>\n The rules target professional and commercial use. Individuals using AI purely for personal purposes are not affected.<\/p>\n Exemptions or lighter regimes also exist for artistic, creative, satirical and fictional works. Text that has undergone genuine human editorial review and for which a person or organisation accepts editorial responsibility is generally exempt from the public-interest text labelling requirement.<\/p>\n Major platforms have begun preparing:<\/p>\n Some industry voices have warned that overlapping labels could create confusion for users, while supporters argue the rules are essential to preserve public trust.<\/p>\n \u201cGenerative AI enables disinformation to be created at unprecedented scale, tailored to specific audiences, and disseminated with remarkable speed,\u201d an EU official said. As it becomes harder to distinguish real from synthetic content, the new labelling requirements aim to protect citizens\u2019 ability to trust what they see, hear and read online.<\/p>\n The EU AI content labelling rules from August 2026 mark one of the first major real-world tests of the bloc\u2019s comprehensive AI Act. Companies that act early will be better placed to avoid fines and maintain user trust.<\/p>\nWhat Must Be Labelled Under the EU AI Act<\/h3>\n
\n
How Companies Can Comply<\/h3>\n
\n
Exemptions and Special Cases<\/h3>\n
Big Tech Already Moving<\/h3>\n
\n
Why These Rules Matter<\/h3>\n