Venezuela, China, Biden and extraterrestrials: the disinformation of AI

A blond television presenter seemingly working for a channel called the “House of News” looks into the camera and asks in English: “How true is it that Venezuela is such a poor country?”

But this man is not a journalist—or even a real human being. He’s an avatar generated by artificial intelligence (AI), a tool that is becoming ever more common in global disinformation.

Subtitled clips from this supposed news channel have appeared on social media with a clear bias towards the Venezuelan government, creating a scandal in the country.

And this is not an isolated case.

“More and more regions are starting to refer to deep fakes and AI-generated audio-visuals as a threat, usually closer to elections,” Shirin Anlen, a media technology expert from the Witness NGO, told AFP.

Last month a US-based research firm identified a fake news outlet called “Wolf News” whose AI anchors pumped out propaganda promoting the interests of the Chinese Communist Party.

A doctored video of US President Joe Biden warning of the threat of an “extraterrestrial nature” went viral last month.

And last year, a video showed Ukraine President Volodymyr Zelensky snorting cocaine while another had US star Eminem attacking Mexico President Andres Manuel Lopez Obrador in a rap song.

Easy to use

The “House of News Espanol” avatars were created by the Synthesia AI program.

Synthesia reported the fake news channel’s content on YouTube, which duly removed it.

“If you’re pretending to be a journalist, or you’re pretending to have breaking news, it’s going to be banned,” Synthesia spokeswoman Laura Morelli told AFP.

However state television channel VTV played a House of News video that boasted about how much money a domestic baseball competition had made, despite no official figures existing.

VTV’s high profile presenter, Barry Cartaya, even claimed the numbers were being talked about “in the United States and a large part of Latin America.”

President Nicolas Maduro has bristled at the suggestion his government is behind the fake videos.

“It is not artificial intelligence, it is popular intelligence,” he said last month. “I am a robot,” he added sarcastically, while trying to imitate one.

‘A lie repeated 1,000 times’

House of News’ two presenters, Noah and Daren, are among Synthesia’s catalogue of 93 avatars—as are the Wolf News duo Alex and Jason.

Avatars were also used in January to create videos of supposed Americans supporting a coup in Burkina Faso.

Meanwhile Facebook’s parent company Meta recently dismantled networks of fake accounts.

One, in Cuba, created fake identities “criticizing members of the opposition,” said Ben Nimmo, Meta’s global threat intelligence lead.

Some used profile photos that were likely generated using artificial intelligence, he said, adding: “This is a tactic that we’ve seen used more and more often by influence operations in the last few years.”

“These types of programs can be used by anyone with basic understanding,” said Eduardo Mosqueira, a professor at the University of A Coruna.

Mosqueira says it is not difficult to identify a fake video, but much tougher to decipher its origin.

“If they have been a tiny bit careful, I imagine it’s impossible,” he said.

And, he adds, it is “increasingly easy to program bots on social media that pretend to be real people and launch messages that you want to get across.”

More often than not, the tactic works.

Mosqueira explains it by citing a quote attributed to Nazi propagandist Joseph Goebbels.

“A lie repeated 1,000 times becomes the truth,” he said.

© 2023 AFP