Latest / AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence / When Your AI Influencer Is Actually a Chatbot
Transcript
- Lucas: So I stumbled onto something this week that genuinely surprised me. There is an Instagram influencer named Aitana López. She's Spanish, has over 200,000 followers, posts fitness content and travel photos, does brand collaborations. She's also completely fictional. She's a computer-generated model run by a Barcelona-based agency. Luna: Wait — people are following a bot and don't know it? Lucas: Some followers know, but many don't. The agency behind Aitana, called The Clueless, says she's easier to manage than a human influencer — no bad behavior, no aging, no scheduling conflicts. And she's profitable. She earns about three thousand euros a month from sponsored posts. Luna: That's actually kind of disturbing. I mean, we've had CGI characters in movies for decades, but this feels different. She's presented as a real person living a real life. Lucas: Exactly. And she's not even the biggest example. Lil Miquela — a virtual influencer created by a startup called Brud — has 2.6 million followers on Instagram. She's been featured in campaigns for Prada, Calvin Klein, Samsung. She's also been 'outed' as CGI multiple times, and yet her audience continues to grow. Luna: So is there any disclosure? Like, do these accounts have to say 'this is an AI'? Lucas: In most places, no. The Federal Trade Commission in the US has guidelines that say endorsements must be honest about material connections, but they haven't specifically addressed synthetic personas yet. Some digital creators voluntarily tag posts as 'virtual influencer' or 'CGI,' but it's not required. And the platform algorithms don't flag them. Luna: That feels like a loophole the size of a stadium. Brands are paying real money to have a fake person promote products to real people who think they're looking at a human. Lucas: And that's where the ethics get really tangled. Let's start with the obvious one: deception. If a follower doesn't know Aitana isn't real, they're forming emotional connections with a fiction. They might comment 'you're so inspiring' and get a reply generated by a language model. Luna: There's a psychological impact there. Especially for younger users who already struggle with comparing themselves to curated images. Now the bar is literally unattainable — it's not even a human. Lucas: Right. Aitana's body proportions, skin, and lighting are engineered to maximize engagement metrics. She never has a bad hair day. She never ages. She's effectively a hyper-optimized fantasy, and that can amplify body image issues. A 2023 study from the UK's Royal Society for Public Health found that two-thirds of young people reported that social media makes their body image worse. Virtual influencers take that dynamic and dial it up. Luna: Then there's the economic angle. Human influencers — especially micro-influencers — earn real income from brand deals. If agencies can just spin up a virtual model for a one-time creation cost, why hire a human? Lucas: Exactly. The agency behind Aitana says their virtual models cost about three hundred dollars to create initially, then a few hundred per month for management. Compare that to a human influencer who might charge five hundred to five thousand dollars per post. The economics are brutal for people trying to build a career. Luna: And that's before we get into cultural appropriation or representation. Aitana is a white, conventionally attractive woman. But agencies could easily create virtual influencers of any ethnicity, gender, or background — and then claim diversity without actually employing diverse people. Lucas: That's a really sharp point. You could have a brand claim 'we work with Black creators' when the creator is a digital asset. It's performative diversity at scale. There was a controversy in 2020 when a virtual influencer named Shudu — a digital Black woman — was promoted as a model for fashion brands, and people asked 'where are the real Black models?' The creator, a white photographer, faced backlash for profiting off a synthetic Black persona. Luna: So what's being done about this? Any regulation on the horizon? Lucas: Brazil actually passed a law in 2024 that requires influencers — including virtual ones — to clearly label paid content and disclose if the persona is ai generated. The fine for non-compliance can be up to ten percent of the influencer's revenue. In the EU, the Digital Services Act is starting to address synthetic media, but it's still vague on individual influencer accounts. Luna: But enforcement is the hard part. How do you police millions of accounts? And what about platforms like TikTok where virtual influencers are exploding? Lucas: TikTok has a policy that 'synthetic media' that shows realistic scenes must be labeled, but enforcement seems inconsistent. And the platforms have an incentive not to crack down — these accounts drive engagement. Lil Miquela's content generates millions of likes. The algorithm loves her. Luna: I wonder if there's also a liability question. If a virtual influencer promotes a harmful product — say, a questionable diet supplement — who's responsible? The agency that created her? The brand? The platform? Lucas: That's a legal gray area. A human influencer can be held liable for false claims. But a virtual influencer is essentially a copyright — there's no person to sue. The agency could argue they're just providing a character, not endorsing anything. It's a potential accountability black hole. Luna: And that's exactly the kind of gap that bad actors could exploit. Imagine a virtual influencer with millions of followers slowly being used to push political content or disinformation. Lucas: That's not hypothetical. During the 2024 US election, researchers identified several virtual accounts that were posting political content, though it was hard to tell if they were genuine efforts or just engagement bait. The line between marketing, propaganda, and entertainment is completely blurred when the source is synthetic. Luna: So what's the solution? Should we just ban virtual influencers? Lucas: I don't think a ban is realistic or even desirable. There are legitimate creative uses — virtual brand mascots, educational characters, artistic projects. But I think we need three things: clear labeling that's obvious and machine-readable, liability that falls on the creating entity, and platform accountability for amplification. Luna: Speaking of accountability — and this ties back to what we're talking about trusting what we see online — it's part of why I'm glad this show stays independent. We don't have to answer to advertisers who might want us to soft-pedal a topic. And that independence is something we can only keep with listener support. Lucas: It's true. We don't run ads, we don't take sponsor money, and that means we can follow these stories wherever they lead without worrying about upsetting a brand. If you value that kind of freedom, you can help keep it going at buy me a coffee dot com slash fexingo. Every contribution, even a few dollars, makes a difference. Luna: Yeah, exactly. And now back to virtual influencers — because I want to dig into one more angle. What about the data these bots collect from their followers? Lucas: Great question. When a follower interacts with a virtual influencer's post or DMs, the agency collects that engagement data — just like any account. But because the persona is automated, they can scale one-to-one conversations in a way a human never could. That's a massive data collection operation wrapped in a friendly face. Luna: So you might think you're sharing something with a person, but you're really feeding a database. That feels like a consent issue. Lucas: It absolutely is. Most platform terms of service don't require a bot to disclose that it's a bot when messaging. So a virtual influencer could send a seemingly personal reply to thousands of people, gathering sentiment data, preferences, even location info — all under the guise of a human connection. Luna: Is there any pushback from the influencer industry itself? I mean, real influencers must see these virtual ones as competition. Lucas: Some are speaking out. A group of human influencers started a campaign called 'Real People, Real Influence' that pushes for transparency. But the agencies creating virtual influencers argue it's just a new creative medium. The Clueless, the agency behind Aitana, says they're not hiding anything — they just don't think disclosure is required by law. Luna: Legally maybe not, but ethically it seems clear. If your audience believes you're human and you let them believe that, it's a lie. Lucas: And that's the core of it. We're moving into a world where the default assumption online might shift from 'this is real' to 'this might be synthetic.' And that erosion of trust affects everything — not just influencers, but news, reviews, recommendations. Luna: It's a huge topic. I feel like we've barely scratched the surface. Lucas: We'll definitely come back to it. For now, I think the takeaway is: if you're following an account that seems too perfect, check the bio, look for disclosure tags, and remember that behind every virtual smile is a line of code — and a company counting your engagement.