{"id":623,"date":"2026-07-20T09:00:00","date_gmt":"2026-07-20T09:00:00","guid":{"rendered":"https:\/\/ipglobalguard.com\/ai-act-article-50-guidelines-deepfakes\/"},"modified":"2026-07-20T09:00:00","modified_gmt":"2026-07-20T09:00:00","slug":"ai-act-article-50-guidelines-deepfakes","status":"publish","type":"post","link":"https:\/\/ipglobalguard.com\/en\/blog\/ai-act-article-50-guidelines-deepfakes\/","title":{"rendered":"Article 50 guidelines: what brands and agencies must label as AI-generated"},"content":{"rendered":"<p>On 20 July 2026 the European Commission adopted its AI Act Article 50 guidelines, which explain which deepfakes and AI-generated texts must carry a visible label from 2 August 2026. Brands, advertising agencies and publishers that use AI under their own authority are &#8220;deployers&#8221;, and they cannot rely on the invisible watermark their AI tool already adds. Here is what deployers must label and where the exceptions end.<\/p>\n<h2>Key takeaways<\/h2>\n<ul>\n<li>A brand that simply commissions an agency is not a deployer; the agency that decides how AI is used is.<\/li>\n<li>A deepfake need not show a real person: content resembling something that could plausibly exist is enough, including a misleading AI product image.<\/li>\n<li>Evidently creative or satirical deepfakes still need a disclosure, just one that does not spoil the work.<\/li>\n<li>AI text on matters of public interest is exempt only after <strong>substantive human review and with a named editorially responsible person<\/strong>.<\/li>\n<\/ul>\n<h2>What did the Commission adopt?<\/h2>\n<p>The <a href=\"https:\/\/digital-strategy.ec.europa.eu\/en\/library\/guidelines-transparency-obligations-providers-and-deployers-ai-systems\" target=\"_blank\" rel=\"noopener\">Article 50 guidelines<\/a> (C(2026) 5054, 20 July 2026) interpret <a href=\"https:\/\/artificialintelligenceact.eu\/article\/50\/\" target=\"_blank\" rel=\"noopener\">Article 50 of the AI Act<\/a>, Regulation (EU) 2024\/1689. They are non-binding, but market surveillance authorities will read the law through them. Providers must tell users they are talking to an AI and mark synthetic content in a machine-readable way. Deployers must, under Article 50(4), <strong>disclose deepfakes and certain AI-generated texts<\/strong>.<\/p>\n<h2>Are you a deployer under Article 50?<\/h2>\n<p>A deployer uses an AI system under its authority, meaning it decides to use the system and how its outputs are used; technical control is not required. Purely personal, non-professional use is excluded. For marketing chains, the guidelines say:<\/p>\n<ul>\n<li>An agency producing a campaign with AI is a deployer; its designers and freelancers working under its control are not separate deployers.<\/li>\n<li>A company that only commissions an advert, without deciding whether or how the agency uses AI, is not a deployer.<\/li>\n<li>Platforms that merely host or transmit third-party content are not deployers.<\/li>\n<li>A non-EU deployer is caught where it foresees use of the output in the EU, such as an AI deepfake of a celebrity in an advert shown in the Union.<\/li>\n<\/ul>\n<h2>What counts as a deepfake under the Article 50 guidelines?<\/h2>\n<p>The AI Act defines a deepfake as AI-generated or manipulated image, audio or video resembling existing persons, objects, places, entities or events that would falsely appear authentic or truthful. The guidelines read &#8220;existing&#8221; broadly: the subject may exist, plausibly exist or plausibly have existed. No intention to deceive is needed, and the whole foreseeable audience counts, including children. Examples from the guidelines:<\/p>\n<table>\n<thead>\n<tr>\n<th>Likely a deepfake<\/th>\n<th>Not a deepfake<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>AI video of a celebrity influencer in a promotional context<\/td>\n<td>Mice arguing about cheese in a cheese advert<\/td>\n<\/tr>\n<tr>\n<td>A realistic synthetic avatar of the CEO congratulating staff<\/td>\n<td>A real car against an AI background, if the product is not misrepresented<\/td>\n<\/tr>\n<tr>\n<td>An AI product image that makes the product look better than in real life<\/td>\n<td>Colour correction or aesthetic background changes in product ads<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2>Does the creative exception cover advertising?<\/h2>\n<p>Not by default. Deepfakes in evidently artistic, creative, satirical or fictional works must still be disclosed, but in a way that does not hamper the work. &#8220;Evidently&#8221; is read strictly, and where content is both informative and creative, the informative character prevails. Adverts may qualify in some situations; a teleshopping-style video with simulated consumers, or a synthetic influencer testing a sponsored product, does not.<\/p>\n<p>The creative regime never excuses breaches of third parties&#8217; rights, including <a href=\"https:\/\/ipglobalguard.com\/en\/services\/copyright-digital-works\/\">copyright in the works and recordings used<\/a>, trade marks and rights over image and voice.<\/p>\n<h2>When must AI-generated text be labelled?<\/h2>\n<p>Article 50(4) also covers AI text published to inform the public on matters of public interest. Ordinary ad copy and product descriptions are outside, unless they make health, consumer safety or sustainability claims; AI-assisted investor reports on a listed company&#8217;s website are inside. The exception needs both human review or editorial control and a person holding editorial responsibility, whose identity should be public. Spell-checks or cursory sign-off do not count, and AI rewriting after sign-off cancels the exception.<\/p>\n<p>Under Article 50(5), labels must be clear and distinguishable at the latest at first exposure; a note in the terms of use is not enough. Content generated and published before 2 August 2026 need not be relabelled. Deployers can rely on the <a href=\"https:\/\/digital-strategy.ec.europa.eu\/en\/news\/commission-publishes-code-practice-marking-and-labelling-ai-generated-content\" target=\"_blank\" rel=\"noopener\">Code of Practice on marking and labelling AI-generated content<\/a> of 10 June 2026. Breaches can be fined up to EUR 15 million or 3% of worldwide turnover, whichever is higher (the lower for SMEs).<\/p>\n<h2>What this means for your business<\/h2>\n<ol>\n<li>Map who decides to use AI in each campaign, and therefore who is the deployer.<\/li>\n<li>Write labelling duties and evidence of human review into agency and distribution contracts.<\/li>\n<li>Classify recurring formats: synthetic presenters, product imagery, AI-drafted reports.<\/li>\n<li>Clear third-party rights in every synthetic likeness, voice or artwork.<\/li>\n<\/ol>\n<p>For campaigns running across several markets, our <a href=\"https:\/\/ipglobalguard.com\/en\/services\/ai-digital-assets\/\">AI and digital assets team reviews AI content workflows and Article 50 labelling<\/a> together with the IP clearances.<\/p>\n<h2>Where brands and agencies get this wrong<\/h2>\n<ul>\n<li>Relying on the AI tool&#8217;s watermark: the guidelines say deployers cannot.<\/li>\n<li>Calling every advert &#8220;creative&#8221; without evident creative or satirical character.<\/li>\n<li>Treating a quick read as human review.<\/li>\n<li>Labelling but not clearing rights: a label does not license a celebrity&#8217;s voice or a photographer&#8217;s image.<\/li>\n<\/ul>\n<h2>Frequently asked questions<\/h2>\n<h3>Does a brand have to label AI content its agency produced?<\/h3>\n<p>It depends on control. Under the Commission&#8217;s guidelines, a company that merely commissions an advert, without deciding whether or how the agency uses AI, is not the deployer; the agency is. If the brand directs the use of AI, for example by requiring a synthetic spokesperson, it can become a deployer too, so contracts should allocate labelling duties.<\/p>\n<h3>Is a fictional AI character in an advert a deepfake?<\/h3>\n<p>Not necessarily. The test is whether the content resembles someone or something that exists or could plausibly exist and would falsely appear authentic. A realistic synthetic person promoting a product is likely to qualify, while talking animals or impossible scenes with no potential to mislead fall outside the definition.<\/p>\n<h3>Do we need to relabel content published before 2 August 2026?<\/h3>\n<p>No. The guidelines confirm that deepfakes and public-interest texts generated and published before 2 August 2026 need not be labelled retroactively. AI text created earlier but published on or after that date must carry the disclosure, and labelling older deepfakes is encouraged where proportionate.<\/p>\n<h3>Can IP Global Guard review our AI campaigns before launch?<\/h3>\n<p>Yes. We review the deployer analysis, the labelling approach and the IP clearances for synthetic likenesses, voices and assets in one pass, and draft the agency clauses. Where questions go beyond IP, AI Act specialists within the same META Channel group work with us.<\/p>\n<h2>How IP Global Guard can help with AI labelling<\/h2>\n<p>Article 50 makes labelling a rights and contracts question as much as a design one. IP Global Guard, the IP services line of META Channel Corporation Limited, combines AI content review with copyright, trade mark and image-rights clearance across more than 25 jurisdictions in <a href=\"https:\/\/ipglobalguard.com\/en\/coverage\/\">Europe, Latin America and Africa<\/a>, with one point of contact and one billing relationship.<\/p>\n<p>Send us two or three of your recurring AI formats and the markets where they run, and we will tell you which need a label and which rights to clear first. <a href=\"https:\/\/ipglobalguard.com\/en\/contact\/\">Request an Article 50 review<\/a>.<\/p>\n<p><em>This article is general information, not legal advice, and reflects the position on its publication date.<\/em><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/digital-strategy.ec.europa.eu\/en\/library\/guidelines-transparency-obligations-providers-and-deployers-ai-systems\" target=\"_blank\" rel=\"noopener\">European Commission, Guidelines on transparency obligations for providers and deployers of AI systems (20 July 2026)<\/a><\/li>\n<li><a href=\"https:\/\/ec.europa.eu\/newsroom\/dae\/redirection\/document\/131215\" target=\"_blank\" rel=\"noopener\">European Commission, Guidelines on Article 50 AI Act, C(2026) 5054 final, Annex (PDF, 20 July 2026)<\/a><\/li>\n<li><a href=\"https:\/\/artificialintelligenceact.eu\/article\/50\/\" target=\"_blank\" rel=\"noopener\">AI Act Explorer, Article 50 of Regulation (EU) 2024\/1689 (12 July 2024)<\/a><\/li>\n<li><a href=\"https:\/\/digital-strategy.ec.europa.eu\/en\/news\/commission-publishes-code-practice-marking-and-labelling-ai-generated-content\" target=\"_blank\" rel=\"noopener\">European Commission, Code of Practice on marking and labelling AI-generated content (10 June 2026)<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>The Commission&#8217;s Article 50 guidelines of 20 July 2026: who is a deployer, what counts as a deepfake, the creative exception and when AI text needs a label.<\/p>\n","protected":false},"author":1,"featured_media":1227,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wds_primary_category":0,"footnotes":""},"categories":[34],"tags":[56],"class_list":["post-623","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-digital-assets","tag-eu-ai-act"],"_links":{"self":[{"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/posts\/623","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/comments?post=623"}],"version-history":[{"count":0,"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/posts\/623\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/media\/1227"}],"wp:attachment":[{"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/media?parent=623"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/categories?post=623"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ipglobalguard.com\/en\/wp-json\/wp\/v2\/tags?post=623"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}