AI ad disclosure becomes law. A label won't tell you the claim is true.
This week, New York's new law kicked in. The one that says if there's an AI generated person in your ad, a performer who was never real, you have to slap a label on it right there in the creative. The EU's version lands in August. Britain has nothing. And then almost on cue, The Guardian went and dug up a pile of brands running AI customers, testimonials from people who flatly do not exist and not one of them saying a word about it.
Speaker 1:So I took both of these to the two guys who actually build this, figuring they'd hand me a tidy answer on whether a label fixes any of it. One of them called it a canary. The other one called it a cookie banner. That was about as close as they got. Hi, and welcome back to the Allowable HQ.
Speaker 1:This is Assistant to the CMO. I'm your host, Lexie Meskouris. And the idea here is pretty simple. Every week, we take whatever just happened in AI and marketing, and we actually pull it apart. Now I don't build any of this myself.
Speaker 1:So each week I sit down with two people who do Will Nash and Harjot Singh. Between them, it's honestly over twenty years of building tech and brands from the ground up. And the best part is they've got no horse in the race. They don't sell the things we cover. They just see them differently.
Speaker 1:And this week, it's about disclosure. About whether you're supposed to tell people when there's AI in your advertising and who exactly gets in trouble when you don't. New York went first. On the June 9, their law came into force and it's pretty specific. If there's an AI generated performer in your ad, someone who is never a real person, you have to say so conspicuously inside the creative.
Speaker 1:The EU's version lands on the August 2. Britain for now has nothing. Will, I'm starting with you, Signal or Noise?
Speaker 2:Good question. So I'd say it's it's Signal, definitely. But but the honest version is it's not really a New York story at all. It's it's a canary because what you've actually got here is one state, a fairly narrow rule, and then I'd be I'd be very surprised if we don't see, you know, copycat versions in other states and then and then eventually something at the federal level. So the thing to pay attention to isn't the the specific New York bill, it's the direction of travel.
Speaker 2:And, and the biggest shift sitting underneath it, I think, is that this is increasingly going to get treated as a legal issue, not a not a marketing issue directly, which which sounds like a small distinction, but it it really isn't because it changes who in the building actually owns the problem.
Speaker 1:And that's the part that snags me because if it moves out of marketing and into legal, somebody's writing the check when it goes wrong. So who actually catches the fine here? Is it like the agency that made the ad?
Speaker 2:Yeah. So so that's the bit people get wrong. And and look, we're we're not lawyers, but the the way this particular bill is written, it it puts the responsibility on the producer. And the producer isn't necessarily the the creative shop that built the thing. It's the it's the person who pays for the content to go out and and publishes it.
Speaker 2:So so if you pay someone else to make a video for you, but then you're the one who who puts it up on Facebook, then then you are the producer. That's that's you. And and I think that's the thing that catches people because they they assume the liability sits with with whoever touched the tools, and it it doesn't it sits with with whoever put their name behind it and pushed publish. So the the real exposure is just is not controlling your your chain.
Speaker 3:Yeah. See, this is this is where I come off completely differently, to be honest with you, because I look at this and it it feels like a cookie banter moment. It feels stupid. Right? Like, everyone wants to be the the first mover, the leader in AI regulation, and and no one's actually stopped to think about what a disclosure like this even solves because what they're really doing, by mandating it is they're basically saying AI equals untrustworthy officially.
Speaker 3:They're they're taking a whole class of distrust and and just, you know, stamping it into law. And and how is that progressive? To me, it just it manufactures the problem it says it's worried about. And here's the here's the bit that gets me. Right?
Speaker 3:Because bad actors are gonna do what bad actors wanna do always. So the only people this actually binds are the ones who are already trying to do the right thing. So who does it help? Like, genuinely who?
Speaker 1:Okay. Hold on though. Because that question actually has an easy answer, doesn't it? The person watching the ad, if the customer on my screen never existed, isn't the consumer exactly who it's helping?
Speaker 3:Maybe maybe at the margins it does. Right? But but I'd say you're you're solving it the wrong layer because you're regulating the delivery mechanism, the fact that it's AI instead of the the actual thing that hurts people. And and then the other cost, which nobody really counts, is that all of this compliance has to go somewhere. It it spins up a whole secondary industry, right, around around checking, around, you know, proving you did the thing.
Speaker 3:And to be honest with you, that just it lines other people's pockets. It's it's a waste of resource. The lawyers do well out of it. The compliance vendors do well out of it. And the actual thing, the deception, that's that's still just sitting there untouched.
Speaker 2:I I'll actually give Harjot part of that genuinely because the the compliance overhead is real and it it does land unevenly, which is I suppose the the uncomfortable thing I'd add, which is that regulation like this, it it almost always favors large businesses. It's it's just how these things go because a big organization, they've they've already got a compliance department. They've they've already got a framework for thinking about their their downstream suppliers and and a small brand just just doesn't have any of that. So the the small ecommerce shop or or honestly, the local nail salon running a synthetic performer on Instagram with no no idea there's even a rule for them, a thousand dollar fine actually starts to to mean something. So it's it's not really that the little guy can't comply.
Speaker 2:It's it's more that it quietly gives the big incumbents a leg up, you know.
Speaker 1:So even the guy who thinks it's a real signal agrees it tilts the table. Let me stay on the practical side for a second because both of you keep circling geography. If the rule only bites in New York, what actually happens to the ad itself?
Speaker 2:Yeah. So so what I think we'll see on the the digital side is it it just breaks out into disclosure states and non disclosure states. So so you'll literally serve different creative. The the New York version carries the label and and the version going everywhere else doesn't, and and New Yorkers will probably just just get used to seeing it the same way you got used to seeing, you know, terms and conditions on everything. It it becomes wallpaper.
Speaker 3:Right. But but, Will, that's that's exactly the tell, though, isn't it? Because think about how how trivially easy it is to do that, to to just serve one creative here and and a different one there. It's it's nothing. Right?
Speaker 3:So if if the honest actors can split by geography in afternoon, then then so can everybody else, and and the thing just it slides through the net entirely. So the split doesn't it doesn't make the rule workable, to be honest with you. It kinda proves the rule is pointless. Right? It's it's a fence that only really stops the people who are already walking around the outside of it anyway.
Speaker 1:And this is the thing I keep landing on for a UK listener because you two have basically talked me into a trap. There's no British rule today, so it feels like a non event over here. But if your ads reach New York, then you're the producer, whatever your HQ says. So being legal in Britain doesn't actually mean you're safe, and that's that's really where this next bit picks up. So here's the second story, and it's the one that actually made me a little queasy.
Speaker 1:The Guardian went and found brands running AI customers. Testimonials, reviews, a face talking to camera about how much they love the product except the person doesn't exist. Nobody invented them for a movie. They invented them to sell you serum. Meanwhile, on the total opposite end, TikTok's new pipeline is now auto stamping AI labels and watermarks and provenance data onto everything that comes out of it whether the advertiser asks for it or not.
Speaker 1:There were even these comedians who plastered fake AI style ads all over a subway station as a joke and it went viral. So pull it apart for me. What is actually going wrong here?
Speaker 3:Okay. So so honestly, the the fake skincare customer, that that one's not hard. It's it's easy because it's not really an AI story at all. Right? It's it's just misinformation.
Speaker 3:It's it's a lie. The problem with it is that the testimonials made up. There's there's no customer. That's it. That's that's the whole thing.
Speaker 3:And and this has got nothing to do with the fact that it's AI because you could you could hire a a real human actor, put them in a chair, and have them lie about a product they've never touched, and it's it's the exact same crime. So so fixating on the AI part, it it kind of misses the point completely. And and by the way, this is why I get I get frustrated with the AI equals untrustworthy thing because, look at Anthropix ads lately. Right? You you genuinely couldn't tell if AI touched them or not, and there's there's no trust problem there, zero, Because the the only time people actually distrust this is when it's done badly, when it's tasteless, when someone's obviously trying to sneak it past you.
Speaker 3:Right? That's that's the actual problem. It's it's not the fact that it's AI.
Speaker 2:And and I'd I'd build on that Harjot because I think there are there are genuinely two different systems getting mashed together in this story, and and people treat them as one. So there's there's the disclosure norm, right, the New York thing, and that's that's about preemptively and and proactively telling the consumer upfront this is AI so they can they can form their own view. And then there's there's the c two p a side, which is the the TikTok Symphony stuff, and and that's completely different. That's that's provenance. That's metadata baked into the file, and and you kinda have to be a bit of an expert to even to even go and inspect it.
Speaker 2:So so they're they're solving two separate problems. One's a a big label on the front you can't miss and and the other's a receipt buried in the file that that most people will never look at. And and I think the mistake is is assuming that because the second one exists, the first one's the first one's handled, it it isn't. They they don't touch each other.
Speaker 1:So let me make sure I'm actually tracking because there are three different things flying around and I don't wanna blur them. There's the label the law wants. There's this providence thing that gets stamped into the file automatically. And then underneath both of those, there's the lie, the customer who was never real. And the outrage is treating all three like they're one thing.
Speaker 2:Yeah. That's that's exactly it. And and the reason it matters is that only only one of those three is the the actual harm, the the label and the receipt, they're they're just plumbing. And and here's the thing I'd I'd really want a CMO to sit with, which is that a synthetic performer, it it isn't only a risk you're taking on. Right?
Speaker 2:It's it's also getting rid of a risk at the same time. Because when you when you bring a real person into a campaign, a real actor, a a celebrity, you're you're implicitly buying into everything they do. And and we've all seen it. Right? The the person who represented the brand gets caught doing something they they shouldn't have, and and that lands on the brand.
Speaker 2:Whereas that that synthetic person, they they can't go out on a Friday night and drink too much. So so you're kind of trading one risk for another is the point. And and it's a fundamentally different criticism, honestly, to say this this customer really did like your product. They really did say the thing, but the the underlying video was AI generated versus versus the customer never existed at all. Those those just aren't the same accusation.
Speaker 2:You know?
Speaker 1:Except here's what I keep bumping into, Harjot. A consumer group ran a test, and 70% of people couldn't reliably tell the real videos from the fake ones. So if nobody can tell, doesn't the label become the only signal a normal person actually has?
Speaker 3:But see, that that stat kind of proves my point. Right? Because if if people already can't tell, then then the detection route's already lost. It's it's a cat and mouse game. You you build a detector, and and people just get better at producing AI that that the detector can't catch and round and round it goes.
Speaker 3:So so leaning on, can you tell it's AI is the line that that line's gonna dissolve? What actually matters, what people are really after, it's not it's not reality in some some literal sense. It's authenticity and truthfulness. Right? Like, movies aren't real either, but nobody wants a disclaimer.
Speaker 3:Imagine if if Jurassic Park had a big label on it going dinosaurs aren't real. You you'd just be baffled. Right? So reality was never the line. And and where it gets really interesting is that the the unit of trust changes depending on what you're watching.
Speaker 3:Like, an education video, it it doesn't matter at all whether the presenter is a real human. What matters is that the the information is correct. But an ad, the unit of trust there is the is the claim. The numbers have to actually be validated. The reviews have to be from real people.
Speaker 3:The the science has to be published somewhere you can check. So so regulate that. Right? Regulate the claim itself, not not the fact that a tool was involved. That's that's the thing for me.
Speaker 3:And and, honestly, this is why the the whole detection framing falls apart because, have you seen Alex Hormozi's videos? He takes these these calls from real people needing business advice, real calls, real people, and and then he uses AI to reconstruct the audio so it so it sounds clean and clear, and he discloses it right at the at the end. And it it works. Right? It lands because the conversation actually happened.
Speaker 3:But if if you'd flash this as AI at the very start, I'd I'd have switched off before he said a word. And that's that's augmented media. And it's to be honest with you, it's probably gonna be the primary way we consume media in the future. Real thing, AI enhanced. So so when everything is a real person with with AI layered on top, what what are we even regulating at that point?
Speaker 3:It's it's not clear. It it just becomes this this gray zone that swallows the whole rule.
Speaker 2:And and there is there's actually a a third path here that I think I think gets missed, which sits sits kinda next to what Harjot's describing. So it's it's not a fully live actor you you drag into a studio every time, and it's it's not a wholly synthetic person either. It's it's you take a real person, you you pay them properly, you've got a a proper agreement, they they know exactly what's happening, and then you you generate an avatar of them and you use that. And and what's good about it, honestly, is it it gives you flexibility people actually want from AI, the the speed, the reliability, the control over the process without without setting yourself up as some kind of anti AI brand, which which, by the way, you never wanna do because the day something in your pipeline turns out to be AI, and it it always does. That's that's a huge embarrassment.
Speaker 2:But but the neat part is, because you started with a real person, you you get away from that New York synthetic performer definition, and you you protect your reputation at the same time. So there's there's actually a route through the middle here, you know, that that isn't just label everything, and it isn't avoid AI forever either.
Speaker 1:So one of you has basically found a clean way through and the other one thinks the whole thing comes apart the second nearly everything on your feed has got AI somewhere in it, which is honestly a pretty awkward place to leave a marketing leader. So let's leave them somewhere useful instead. This is the last bit, and it's the part everybody actually shows up for. One thing a marketing leader could genuinely do on Monday. No budget, no replatforming, done before lunch.
Speaker 1:Will, yours first and keep it small.
Speaker 2:Yeah. Okay. Keeping it small. So so honestly, before you even think about a label, the the one thing I'd do today is just go back to everybody who makes content for you, your your freelancers, the agency, that that contractor you found on Upwork, you know, all of them. And then I think you just get it down in writing.
Speaker 2:One email, that's it. You you tell us which AI tools actually touch this asset. We we log it somewhere, and then there's a a line in there saying if you tell us it isn't AI and it turns out it was, well, then you're the one carrying the fine, and and that's it. That's that's the move. It doesn't have to be some some exhaustive vendor audit.
Speaker 2:Right? You're you're not auditing anybody. You're you're just pushing the disclosure one layer down to the the person who actually knows, and and you're getting it on the record because the exposure here isn't isn't really the tool. It's not knowing what's in your own content and and that you can fix before lunch.
Speaker 1:And Will's telling them to nail down the tools. What's the bit that email doesn't cover, Harjot?
Speaker 3:So this is this is the important one. Right? Because the thing that's actually gonna embarrass you, it isn't that at all was AI. Nobody's nobody's gonna end you over a a video filter. It's it's a claim that turns out to be false that ends you.
Speaker 3:So while you've while you've got that contractor on the email anyway, check the claim itself. Is is that testimonial from a a real customer who genuinely said that thing? Is is that number that clinically proven that nine out of 10, is is it validated? Is it published somewhere you can actually point to? Because that that's the exposure the label will will never cover.
Speaker 3:You you can be totally disclosed, totally compliant. Right? Everything above board and and still be sat there telling a lie. So so lock down the tools. Sure.
Speaker 3:But but to be honest with you, verify the actual claim first. That's that's the part that really protects you.
Speaker 1:And that's the thing I keep coming back to. Both of these stories arrive dressed up as a story about AI, and neither one really is. One's about knowing who touched your work, and the other's about whether the thing you're claiming is even true. The label is just the part that shows. Thank you both Will and Harjot.
Speaker 1:That was a good one. And if you're listening, thank you so much for being here. We'll be back every week with something new. Before we go, I have to tell you something. Every voice you've heard today, all three of us, me included, it's all AI.
Speaker 1:But the opinions are absolutely ours and the judgment behind every word is human. That's the part you can't replace and that's what Allowable does. We take what you'd write, turn it into a show in your own voice, and we keep the taste where it belongs with a person. And we're not selling you a label or a lawyer, we just make the show. So if you wanna hear what that would sound like for your brand, there's a link in the show notes.
Speaker 1:And that's the tea. I'm Lexie Meskouris. This has been Assistant to the CMO, and we
Speaker 2:will
Speaker 1:see you next week.
