A Deepfake Landed in Slovakia's Silent Hours
Two days before Slovakia's September 2023 election, a fabricated audio clip purporting to catch a liberal party leader discussing how to rig the vote spread across social media during a legally mandated pre-election blackout. Fact-checkers identified the recording as manipulated using artificial intelligence, in one of Europe's earliest tests of a deepfake in a live campaign.
On 28 September 2023, two days before Slovaks went to the polls, an audio clip began circulating on Facebook, Instagram, Telegram and chain emails. It purported to record a conversation between Michal Šimečka, the leader of the liberal Progressive Slovakia party, and Monika Tódová, a journalist with the independent newspaper Denník N, discussing how to fix the coming election, according to the International Press Institute. The exchange, in which the two appeared to weigh buying votes from the Roma minority and manipulating the count, never happened.
The clip surfaced during a 48-hour moratorium before polls opened, a silence period in which Slovak media outlets and politicians are meant to refrain from campaigning. Because of those rules, the fabricated recording proved difficult to publicly debunk in the hours when it spread most widely, The Dial reported. On Facebook alone more than 100,000 people saw the recording, and at least another 30,000 on Instagram, according to the International Press Institute.
Fact-checkers flag an AI forgery
The fact-checking department of the news agency AFP examined the clip and said it showed signs of having been manipulated using artificial intelligence, the International Press Institute reported. Both Šimečka and Denník N denied the conversation had taken place. Matúš Kostolný, the editor-in-chief of Denník N, wrote that the file was a deepfake that “speaks in the voice of my colleague Monika Tódová in made-up sentences,” according to the same account.
Press-freedom groups described the episode as one of the first known cases in the European Union of AI being used to imitate the likeness of a journalist. Tódová filed a criminal complaint over the recording, the International Press Institute reported.
A gap in the platforms' rules
The clip also exposed a hole in how social networks police synthetic media. Meta's manipulated-media policy at the time applied only to faked videos in which a person is edited to say words they never spoke, not to audio, Wired reported. An audio-only forgery therefore fell outside the rule most likely to catch it.
What the case did, and did not, prove
Progressive Slovakia lost the election to the party of former prime minister Robert Fico. Researchers have cautioned against treating the deepfake as the decisive factor. In an analysis for the Harvard Kennedy School's Misinformation Review, scholars argued that the clip landed in a country already marked by low institutional trust and years of influence operations, and that “the deepfake's impact must be understood in the context of these long-term influence operations,” the review said.
What the Slovak case did establish, analysts across those accounts agreed, was the timing problem: a convincing forgery, released into an information vacuum when rules or norms keep the press and candidates quiet, can travel far before it can be corrected.