Tightened regulations joined early-mover New York when they took effect this weekend. But what do they mean right now for advertisers? And are tighter global standards now an inevitability?<\/p>\n
New AI disclosure laws are landing worldwide – but what do they mean for marketers? (Unsplash)<\/p>\n
Last weekend, Article 50 of the EU\u2019s landmark AI Act came into force.<\/p>\n
Under it, those who make and use certain kinds of AI systems now have new and expanded transparency obligations. Now, they\u2019ll have to inform users when they\u2019re interacting with an AI system, or when they\u2019re exposing to deepfakes, emotion recognition, or biometric categorization tools. In other words, artificially generated content designed to look real must be labeled as such.<\/p>\n
On the same day that Article 50 came into force, it was joined by the California AI Transparency Act, which imposes disclosure requirements on, among others, popular generative AI systems. It requires those providers to offer free AI detection tools, include an option for \u201cmanifest disclosure\u201d for content generated by their software and embed \u201clatent disclosure\u201d that makes AI content identifiable as such, even when licensed to third parties.<\/p>\n
Those new laws follow early movers like the state of New York, whose synthetic performer disclosure law came into force in June, ahead of the more sweeping Responsible AI Safety and Education Act, expected to take effect on the first day of 2027.<\/p>\n
Laws elsewhere in the world, combined with pressure from brands<\/a> and trade groups<\/a>, together amount to what one ad agency exec \u2013 Redscout chief executive Michael Keany \u2013 today calls \u201c2026, the year of disclosure by default.\u201d<\/p>\n What has changed for advertisers?<\/p>\n Under the EU\u2019s Article 50, the crux of the matter is that advertisers (among other parties) now have to disclose when AI has generated or manipulated depictions of people, places or objects that appear real. Failure to do so can result in fines of up to \u20ac15m or up to 3% of global annual revenue.<\/p>\n Provisions and punishments will vary across jurisdictions, but Gabrielle Robitaille, director of policy and AI & privacy lead at trade body the World Federation of Advertisers (WFA), says that it\u2019s clear we\u2019ve reached a tipping point, following which global lawmakers and other standards-setters will not be slow to follow the EU, New York and California. Robitaille compares the moment we\u2019re in now to GDPR in 2018: \u201cThere was a legal framework, but many of the practical questions about implementation still had to be worked through. We\u2019re now at a similar inflection point.\u201d<\/p>\n The organizations that the WFA represents \u2013 brand-side marketing departments \u2013 have been clamoring for a level of \u201ccertainty,\u201d Robitaille says, with clear, enforceable rules, \u201ca board-level issue.\u201d<\/p>\n But that\u2019s not to say the matter is now settled. The New York synthetic personas statute has faced criticism for its specificity and enforceability. Lawyers will begin to develop a clearer sense of EU and California laws as they are enforced. As such, Robitaille says that \u201cthe legal certainty that global brands need \u2026 won\u2019t happen overnight, but will evolve over the next few years.\u201d She adds: \u201cThe next few weeks and months will be critical as regulators publish guidance, enforcement begins and industry develops common practices.\u201d<\/p>\n \n Want to go deeper? Ask The Drum<\/p>\n How is the industry reacting?<\/p>\n According to WFA research, 82% of multinational advertisers believe transparency is essential to protect their brands\u2019 reputation \u2013 and, even before this latest raft of regulation landed, 67% had already developed internal rules for disclosure around AI-generated marketing creative.<\/p>\n On the agency side, too, further clarity on regulation is being welcomed by many insofar as it starts to clear up the uncertainty that has swirled around disclosure for at least a few years. Redscout\u2019s Michael Keany says that while current regulation \u201cmay be a little behind the jagged frontier,\u201d it\u2019s still useful. \u201cIf you look at some of those components that they\u2019re going after, it is managing things that, if left unregulated, could lead to really unpleasant things\u2026 I like good regulation because it provides, if nothing else, health and hygiene around the worst parts of the market.\u201d<\/p>\n For Keany, tightening regulation will only hasten a necessary \u2013 and exciting \u2013 evolution in the industry\u2019s use of AI. \u201cAI round one was about operating efficiencies and everybody loving LLMs for research. Round two is about authenticity and trust \u2013 and some economic returns. There\u2019s a tiny bit of tension between perhaps the office of the CFO looking to generate operating efficiencies and the office of the CMO trying to figure out how to generate more demand in a more cacophonous, inauthentic, sloppy world.\u201d<\/p>\n