From apparently conspiring to steal elections to abusing staff or performing embarrassing dance moves, the last few years have seen plenty of fake content made about politicians intended to malign its targets and mislead the public.<\/p>\n
Now EU rules aim to stem the flood of this deceptive content: starting from Sunday, artificially generated images, audio and text designed to look authentic must be labelled.<\/p>\n
The rules, devised under the EU\u2019s landmark AI Act<\/a>, mean companies must ensure people know when they are interacting with an AI-powered chatbot or viewing a manipulated image or AI-generated text.<\/p>\n Deepfakes have included fake audio of the Slovakian opposition leader, Michal \u0160ime\u010dka, conspiring with a journalist to rig election. Photograph: Radovan Stoklasa\/Reuters<\/p>\n The rules apply to new AI systems on the EU market from 2 August, while existing ones have an extra four months to comply.<\/p>\n \u201cIt is a matter not only of customer protection, it\u2019s also a matter of democracy protection,\u201d the Green MEP Sergey Lagodinsky, who helped negotiate the AI Act, told the Guardian. \u201cMaking transparent this information is something which we need to preserve our democracy and the authenticity of facts online.\u201d<\/p>\n Under the rules, synthetic text, images, video and audio designed to look truthful must be visibly marked as AI-generated and contain a digital watermark to show the artificial origins of the content.<\/p>\n Texts on matters of public interest must also be labelled as AI-made if there is no human editorial oversight.<\/p>\n Companies are encouraged to label AI-generated content pre-dating the rules, although this is not compulsory. The rules will not apply to users\u2019 personal content and there is an exemption for \u201cevidently artistic\u201d, satirical and fictional works.<\/p>\n Failure to comply could result in fines of up to \u20ac15m (\u00a312.8) or 3% of a company\u2019s worldwide global turnover.<\/p>\n The EU executive has created black-and-white labels anyone can use, although companies can design their own.<\/p>\n EU symbols to be used on content using AI. Illustration: EU<\/p>\n Deepfakes have included audio purporting to reveal<\/a> the Slovakian opposition leader Michal \u0160ime\u010dka conspiring with a journalist to rig elections, mocked-up videos<\/a> of the French president, Emmanuel Macron, dancing in a 1980s nightclub, and fake footage<\/a> of the then UK Labour leader, Keir Starmer, abusing party staff.<\/p>\n The tech industry says it supports the quest to make deepfakes visible, but fears the rules will be interpreted too broadly and overwhelm internet users with banners and labels.<\/p>\n The Computer and Communications Industry Association argues that EU guidelines published in July have expanded the definition of a deepfake far beyond the 2024 AI Act.<\/p>\n \u201cThe label was meant to flag deceptive content,\u201d said CCIA Europe\u2019s AI policy lead, Boniface de Champris. \u201cThat distinction has been removed, so now almost everything gets labelled \u2013 a landscape in an advert is the same as a manipulated political speech. Like cookie banners, once labels are everywhere, users stop noticing them. That doesn\u2019t protect anyone, it just burdens commercial activity across the board.\u201d<\/p>\n He thinks the most visible change for the public will not be on social media, which \u201calready labels most AI content\u201d but \u201cin areas where AI is already used at scale but people don\u2019t know it \u2013 advertising, film, publishing\u201d.<\/p>\n \u201cExpect a lot more labels in sectors not usually associated with AI, and the burden will fall hardest on those least versed in AI regulation,\u201d De Champris said.<\/p>\n Some of the world\u2019s largest tech companies already have policies designed to inform users about AI-generated content.<\/p>\n