Given the proliferation of AI deepfakes in the recent Slovakian election, it's getting harder to tell who's talking. But first...
The Cyber Angle
Something didn't add up in an alleged conversation between Progressive Slovakia's leader, Michal Simecka, and a local journalist that circulated in the run-up to Slovakia's elections on Saturday.
The speech was stilted and their voices flat even as the leader of the country's main pro-European party seemed to slag local voters, discuss buying votes from the Roma minority and joke about child pornography.
If it sounded off, it's because it was. AFP fact checkers concluded the recording was a hoax synthesized by an artificial intelligence tool trained on samples of the speakers' voices. It was one of a handful of fakes that made the rounds on social media, messenger apps and email, including one where a person that sounded like Simecka plotted to jack up beer prices after the elections.
My colleagues Jillian Deutsch and Daniel Hornak detailed the use of disinformation in Slovakia's election, and Olivia Solon documented the role of AI deepfakes.
Simecka denounced the fakes as "colossal, obvious stupidity," and there are no clues who was behind them. It's also impossible to quantify their impact in the election after Progressive Slovakia placed second behind the frontrunner Smer, led by a populist former prime minister who has derided the European Union's sanctions against Russia.
What is clear is that a new era of disinformation is dawning. While experts have long warned about the use of deepfakes to sway voters, AI is now cheap and accessible enough for anyone to try their hands at it.
"Even though the deepfake was technically quite crude --- you could definitely hear that this was not a real person --- this recording spread rapidly," said Daniel Milo, the head of a unit at the Slovak Interior Ministry that fights disinformation. "In one or two years' time, you might not be able to tell the difference."
Rapidly improving technology, coupled with a number of high-profile hacks targeting voter rolls around the world, suggest the problem is only going to get worse.
This year the UK's Electoral Commission revealed a breach that potentially exposed the data of 40 million voters, while Special Council Robert Mueller's investigation into interference in the 2016 US election found that Russian hackers targeted voter data in many states.
Deepfakes containing hate speech or electoral disinformation in Slovakia spread via Meta Platforms Inc.'s Facebook and Instagram, X and Alphabet Inc.'s YouTube, according to a report last week by Reset, a research group that looks at technology's impact on democracy.
More than half of the posts Reset tracked came via Telegram, a messaging app. It didn't review emails, although there were reports of the clips being circulated that way as well. Milo said the platforms' responses were "inconsistent," with some of them deleting or labeling the problematic posts.
The Slovak Interior Ministry is in Meta's government reporting channel, according to Meta spokesperson Ben Walters. "We did not see many escalations from them during the election period but responded in a timely manner to those we did receive," he said.
YouTube has a priority channel with local officials to address flagged content that violates its policies, according to Google spokesperson Ciaran Ward.
Spokespeople for X and Telegram did not respond to requests for comment.
The elections in Slovakia, an eastern European nation of 5.4 million, foreshadow growing divisions between allies regarding assistance to Ukraine as progress in its war with Russia stalls. But more ominously, they also gave a preview of how deepfakes are likely to become a part of political reality.
"This is the first time, as far as I'm aware, that there was a serious attempt to influence the outcome of elections using artificially produced content" to create deepfakes, Milo said. "This is what the future is going to look like."
As if trusting politicians wasn't hard enough already.