Deepfake video targeting Zelensky’s wife linked to Russian disinformation campaign, CNN analysis shows | CNN

London
CNN
 — 

A deepfake video that wrongly suggested the wife of Ukrainian President Volodymyr Zelensky purchased a luxury car during the couple’s visit to Paris last month is likely part of a Russian-linked disinformation network, according to CNN’s analysis and disinformation experts.

The AI-generated video features a non-existent employee of the Bugatti dealership in Paris claiming Olena Zelenska purchased its new model, the Bugatti Tourbillon, for 4.5 million euros (around $4.8 million) on June 7.

The clip has several markings of a deep fake, from cuts in the video, to the strange accent and lip-and-mouth movements, according to Clément Briens, a researcher at cybersecurity company Recorded Future. Yet, it gathered some 18 million views in 24 hours on X, where it was reposted by pro-Russian influencers.

Bugatti said in a statement on Tuesday that its dealership in Paris, operated by Autofficina Parigi, a Car Lovers Group company, was victim of several criminal offenses and “firmly denounced this disinformation campaign.”

“A supposed salesman claiming to belong to the Car Lovers Group and its sports label Schumacher Group, published a video on social networks in which he indicated that the Bugatti Paris dealership sold a vehicle to the Ukrainian presidential couple,” it said, adding that the group “strongly denies both the existence of the transaction and, consequently, the existence of the invoice.”

“The mandatory legal details don’t appear on the invoice, the price of the vehicle is also wrong, the price of the options and their descriptions are inaccurate and inconsistent, the graphics are outdated, and the Car Lovers Group would never have allowed such a document to be issued,” the statement added.

The group also said it has taken legal action “by filing a criminal complaint for forgery, use of forgeries, usurpation of identity and defamation” – but did not specify against whom the complaint was filed.

Clarity, an AI cybersecurity startup that is fighting against deepfakes, told CNN that their review of the video indicated “a high certainty for AI manipulations, mostly around the middle of the face area.”

The video originates from a French-language website, Verite Cachee France (sic), whose content appears to be AI-generated by scraping French media.

A CNN analysis shows Verite Cachee was set up recently, on June 22, 2024, and some of its pages still have prompts with AI to create fake articles at the top of the piece. CNN is not able to find a contact for the website.

The website’s title also doesn’t use accents – the proper spelling would be Vérité Cachée – casting further doubts on its authenticity.

Darren Linvill, a professor at Clemson University’s Media Forensics Hub, told CNN the deep fake and Verite Cachee’s websites bear the hallmarks of a Russian narrative laundering campaign that has been going on since last August, with the aim of undermining Ukraine.

“They typically place some video with a story to tell on YouTube. They then tell that story by layering it on the fake news pages they’ve created, and often also one or more allied web pages. They then integrate the story through social media, starting with real pro-Russian influencers who are part of their network,” he told CNN. “Only minor change with this campaign is that the video doesn’t seem to be on YouTube.”

In a report from last December, Linvill and Patrick Warren, also a professor at Clemson, showed how the disinformation campaign was led by DC Weekly, a website that published a series of AI-generated Russian propaganda and fake stories, such as the debunked claim that Olena Zelenska bought $1.1 million jewellery at Cartier in New York last September.

A report by Recorded Future, a leading cybersecurity company, also identified Veritee Cachee as being part of the same disinformation network.

Given the size and resources of the network, it is likely some Russian support or financing is happening, Clément Briens, a senior threat intelligence analyst at the company, told CNN.

“They scrape articles automatically from a number of sources, using LLMs to introduce specifically political bias to attack Zelensky, Biden and NATO. Then they upload the articles to launder a pro-Russian narrative,” Briens said.

The sheer volume of articles makes it harder to detect when deep fakes are introduced because when someone clicks on the website, they see a backlog of articles that usually pass the initial check, he said.

“Then after someone uploads the video from YouTube or directly into the site, the amplification system gets activated on social media.”

The target is political readers in Europe, with the purpose to erode domestic political support for Ukraine and undermine European leaders that support Ukraine, such as French President Emmanuel Macron, according to the analyst.