Scammers Use AI-Generated Faces of Glapiński and Brzoska in Fake Investment Ads

Ads promising quick profits, allegedly featuring National Bank of Poland (NBP) governor Adam Glapiński or InPost founder Rafał Brzoska, have been flooding Facebook and Instagram for months. Behind some of them aren't photo edits but faces, voices and backstories fully generated by artificial intelligence, designed to lend credibility to fictitious investment offers.

How the fake ads work

The pattern is almost always the same. An ad on Facebook or Instagram shows a familiar face alongside a quote promising extraordinary investment returns, often in cryptocurrency or stocks. Clicking through takes the user to a fake site made to look like a news outlet, where the victim leaves contact details and a phone number, and ultimately money for a financial product that doesn't actually exist.

Such scams used to rely on stolen photos and crude photo edits that were easy to spot. Generative tools have changed the scale of the problem: the face of a fictitious person, video footage with a cloned voice and the whole veneer of credibility can now be produced in minutes, without stealing anyone's likeness at all.

Faces of well-known Poles

Among the most frequently used faces are NBP governor Adam Glapiński, InPost founder Rafał Brzoska together with his wife Omenaa Mensah, businessman Michał Sołowow, as well as celebrities, politicians and journalists. The names and faces of recognizable people are meant to shortcut the viewer's trust.

Rafał Brzoska filed a complaint over the matter with the Urząd Ochrony Danych Osobowych (UODO, Poland's Personal Data Protection Office), demanding that Meta be barred from publishing ads that use his image without consent. The Warsaw Court of Appeals upheld the ruling in his favor, setting a precedent for other public figures whose likenesses end up in fake campaigns.

Why it works

The most troubling finding comes from a study cited by EurekAlert. Participants struggled to tell real faces from synthetic ones, correctly identifying them only 58.4 percent of the time, barely better than random guessing. More importantly, they rated AI-generated faces as more trustworthy than photos of real people.

Paradoxically, faces from newer diffusion models scored highest for trustworthiness, even though they performed worse on realism tests than older GAN models. A generated face doesn't need to be visually perfect to deceive someone, it just needs to feel trustworthy.

Institutional response and scale of losses

CERT Orange Polska listed fake investments among the threats it blocked most often in 2025, showing that the scale of the problem is growing alongside access to cheap generative tools. Since 2023, the Polish police have run the #ZnamTeNumery educational campaign, under which they teamed up with the Puls TV channel and the Fundacja Pod Dębem foundation to produce a controlled demo video promoting a fictitious investment called "Emerytura na maxa" ("Maximum Pension"), featuring actor Henryk Gołębiewski, to show how easily AI can generate convincing ad material.

Common sense is the best defense against online criminals - Polish Police

Meta says it invests in tools to detect fraudulent ads and removes them once reported, but the scale of revenue the platform is estimated to have drawn from such campaigns in 2025 suggests moderation still isn't keeping pace with how quickly generative AI churns out new variants of fake profiles and content.

What this means for Poland

For Polish internet users, this means that simply recognizing a face in an ad is no longer a guarantee of authenticity, even when it's the central bank governor or a well-known entrepreneur. Cybersecurity experts recommend verifying investment offers only through the official channels of the institutions they claim to involve, and treating promises of guaranteed, high, risk-free returns as a clear warning sign.

The legal precedent Rafał Brzoska secured could pave the way for other public figures to force Meta to remove fake ads using their likeness, though the pace at which new generated-face variants appear means each such case addresses a single instance rather than the mechanism as a whole.