The Fake Polexit Influencers Vanished; EU Scrutiny Outlived Their TikTok Run
The AI-generated women who advocated Poland’s withdrawal from the European Union are no longer on TikTok, but the questions raised by their brief appearance did not disappear with the account. The verified record still supports the central finding: the presenters were synthetic, the profile had abruptly changed identity in December 2025, and its operator has not been publicly identified.
The most important subsequent development came on February 23, 2026. In a published European Commission response, Executive Vice-President Henna Virkkunen said the Commission had discussed the relevant account or accounts with TikTok and third parties. The answer did not announce a separate formal proceeding devoted to the Polexit videos; it instead pointed to the Commission’s continuing TikTok investigation concerning election-integrity risks in Romania.
What the December 2025 campaign actually did
The episode centered on a TikTok profile named “Prawilne_Polki.” It presented different young women speaking Polish about Polexit, national sovereignty, migration and domestic politics. These were not disclosed political creators offering personal opinions: the visible speakers did not exist as real people.
A December 30 verification by Demagog documented seven AI-generated videos and reported that the profile adopted its Polish name on December 13, 2025 after previously carrying unrelated English-language material. The videos and profile subsequently disappeared from TikTok.
That account history matters as much as the imperfect faces or voices. A repurposed profile can arrive with age, activity and potentially an existing audience, making a new political identity look less newly manufactured than it is. The campaign therefore combined two familiar creator-economy techniques—an identifiable on-screen persona and short-form repetition—with a concealed break in the account’s history.
The available evidence does not establish who operated the profile. Russian linguistic influence and alignment with narratives favorable to Moscow were discussed publicly, but those clues are not proof of command, financing or origin. Calling the campaign definitively Russian would go beyond the verified record.
Why synthetic creators change political persuasion
A conventional anonymous propaganda page asks users to trust a logo, slogan or institution. A synthetic influencer can simulate something more intimate: eye contact, speech, emotion and the apparent testimony of an ordinary peer. The political message is delivered through the social grammar of creator video even though there is no accountable creator behind it.
This distinction is important because the deception is not limited to whether an image was generated. The fabricated identity implies personal experience, social position and sincerity. A viewer may reasonably interpret a first-person statement as the opinion of a Polish woman when it is actually a script selected by an unknown operator and performed by generated audiovisual media.
Generative tools also separate the persona from the production constraints faced by human creators. Operators can change faces, scripts and delivery without recruiting new speakers or persuading real people to attach their reputations to the message. That makes political influence content easier to vary while leaving responsibility harder to locate.
“AI slop” describes the low-cost, repetitive quality of much generated media, but low production quality does not make a campaign harmless. Visible lip-sync problems or unnatural expressions may alert attentive viewers, yet discovery feeds are consumed quickly and often without context. The relevant risk is the combination of synthetic testimony, rapid distribution and an account designed to resemble an authentic participant in public debate.
Poland asked Brussels to treat the videos as a platform risk
The Polish response moved the episode beyond an isolated moderation dispute. In a digitally signed letter dated December 29, 2025, Deputy Digital Affairs Minister Dariusz Standerski asked Henna Virkkunen to initiate Digital Services Act proceedings concerning TikTok. He also requested information about the content’s reach, removal and measures intended to prevent renewed dissemination.
The letter described the material as showing characteristics of a coordinated campaign, but that wording should not be mistaken for a completed attribution investigation. It was a government request for supervisory action, not a Commission finding that TikTok had broken the law or that a named foreign actor controlled the account.
The Commission’s later answer narrowed what can responsibly be said about the outcome. Brussels acknowledged direct exchanges about the accounts and reiterated that very large platforms must assess and mitigate foreseeable risks to civic discourse and elections. It also referred to recommendations for recognizable labels on AI-generated media, stronger moderation and detection systems.
However, the response did not say that Standerski’s requested Polexit-specific investigation had been opened. The existing formal TikTok proceeding it cited began on December 17, 2024 and concerns risks surrounding Romania’s presidential election, including recommender systems, manipulation of the service and policies governing paid political content. The two matters share regulatory themes, but they are not the same case.
What remains unresolved
The immediate account-level intervention succeeded in one limited sense: the documented profile and videos ceased to be available. Removal does not reveal how widely the clips were recommended, whether the same operator controlled other accounts, which generation tools were used or whether the audience understood that the speakers were fictional.
There is also no public evidence in the reviewed case record that these videos changed opinions about EU membership. Reach, persuasion and electoral effect are different measurements. A video can attract views without changing a vote, while a modestly viewed clip may still normalize a slogan among a particular community.
For viewers, the strongest warning sign is therefore not a malformed hand or awkward mouth movement alone. More useful questions concern provenance: Did the account recently change its name, language or subject? Is there an identifiable person who can be traced outside the clip? Does the speaker provide verifiable experience, or only emotionally polished first-person claims?
For platforms and regulators, the harder issue is distribution rather than generation by itself. Labeling a synthetic presenter gives users important context, but it does not answer whether coordinated accounts are exploiting recommendation systems or disguising political promotion as ordinary creator speech. The Polexit episode matters because it exposed that gap in a compact form: the fake personalities were temporary, while the accountability problem survived their removal.
Also read:
Subscribe to our newsletter
Get the latest Web3, AI, and crypto news delivered straight to your inbox.