Taylor Swift filed three trademark applications last week to lock down her image and voice, including an Eras tour concert photograph and two sound marks for the phrases "Hey, it's Taylor Swift" and "Hey, it's Taylor." The filings landed as the AI detection firm Copyleaks published research identifying a cluster of TikTok ads that use AI-generated versions of Swift, Rihanna, and Kim Kardashian to push what the company describes as potentially fraudulent services. The timing is not subtle.
The fake ads stitch celebrities into familiar interview settings — red carpet step-and-repeats, late-night couches — and dub them with what Copyleaks calls realistic-sounding voices layered under textured filters that hide the seams in the AI-generated visuals. Instead of answering questions, the deepfaked stars pitch a rewards program purportedly run by TikTok.
One ad reworks footage from Swift's October appearance on The Tonight Show Starring Jimmy Fallon. "I was reading about digital behavior this week and came across a testing feature called TikTok Pay," the deepfaked Swift says. "Certain users are being invited to watch videos and submit opinions." She then tells viewers the program is in limited rollout and adds, "If the page opens for you, don't overthink it." The page, of course, opens for everyone.
Key facts
- 01Taylor Swift filed three trademark applications last week covering an Eras tour photo and the phrases 'Hey, it's Taylor Swift' and 'Hey, it's Taylor.'
- 02Copyleaks identified TikTok ads using AI-generated Swift, Rihanna, and Kim Kardashian to promote a fake rewards program called TikTok Pay.
- 03One ad uses manipulated footage from Swift's October Tonight Show Starring Jimmy Fallon appearance.
- 04The scam funnel ends on a page vibe-coded with Lovable, whose branding appears in the URL despite the TikTok logo.
- 05The Consumer Federation of America sued Meta last week over scam ads on Facebook and Instagram.
Click through and the user lands on a third-party site dressed up with TikTok's name and logo but visibly built on Lovable, the AI app-generation platform whose branding appears both on the page and in the URL. From there, users are prompted to hand over their name and personal information. Copyleaks did not pin down the end use of the harvested data, but the playbook — celebrity deepfake, fake program, identity capture — is by now a standard scam template.
“The deepfaked Swift, built from manipulated footage of an October Tonight Show appearance, pitches a fake TikTok Pay program before routing users to a Lovable-built data harvesting page.”— Jaeden Schafer
Swift is one of the most valuable individual brands in entertainment, and the reputational risk of being repeatedly impersonated by scam advertisers is the kind of problem trademarks are designed to address. A registered likeness gives her lawyers a faster lever to pull against platforms hosting the ads, particularly when those platforms are slow to act on impersonation complaints alone.
The legal pressure on platforms is building from other directions too. Last week, the Consumer Federation of America sued Meta, alleging the company misled Facebook and Instagram users about its enforcement against scam ads while profiting from their continued spread. On Monday, the US Federal Trade Commission reported that social media scams have surged overall, with Facebook scams accounting for the highest total dollar losses among platforms tracked.
Criminal liability is also catching up to deepfake harms in narrower contexts. Earlier this month, an Ohio man became the first person convicted under a new federal law criminalizing nonconsensual intimate visual deceptions. That statute does not address commercial impersonation of celebrities, which is exactly the gap Swift's trademark strategy seems designed to cover.
The Lovable detail is the most awkward piece of the story for the AI tooling industry. Vibe-coding platforms market themselves as letting anyone spin up a working web app from a prompt, and that capability cuts both ways: a scammer running a celebrity deepfake funnel no longer needs a developer to build the landing page that captures the data. The branding showing up in the scam URL suggests the operators did not even bother to strip it.
Copyleaks frames the ads as evidence that detection has not kept pace with generation, and the report stops short of estimating how many users have been funneled through the campaign or how much money has been lost. TikTok's own moderation response to the cluster is also unclear from the research. Without those figures, the scope of the harm remains an open question, even if the mechanics are well documented.
What is clear is that the celebrity deepfake ad has graduated from novelty to infrastructure. The pieces — voice cloning, video manipulation, AI-built landing pages, programmatic ad buying on a major social platform — are all cheap, off-the-shelf, and composable. Swift's trademark filings are a rational response from someone with the resources to litigate, but they are also a tacit admission that the platforms hosting the fraud are not going to solve it on their own. The next phase of this fight will be fought in court filings, not content moderation queues.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
