Authentica Report: Disinformation Campaign Targeting Ukraine, July 2024

On June 7, 2024, a deepfake video emerged online, depicting a fake luxury car dealership employee making a shocking claim about Ukraine’s First Lady. The fictional employee, Jacques Bertin, falsely claimed that Ukrainian President Volodymyr Zelenskyy’s wife, Olena Zelenska, had purchased a Bugatti Tourbillon for 4.5 million euros. The video release included a fabricated invoice to add a layer of credibility. Within 24 hours, the video amassed around 18 million views on X, bolstered by pro-Russian influencers. Zelenska had visited Paris earlier in June with her husband to commemorate the 80th anniversary of D-Day. Subsequent investigations revealed that the video was part of a Russian disinformation campaign. 

What Happened?

The video features an individual sitting in a car parked on the side of the street. The individual is speaking French, but the video features a translated transcript in English subtitles.

Translated Transcript:

“Hello everybody! As you know, a few days ago Bugatti Automobiles introduced its new car model Bugatti Tourbillon…

Now that the model is no longer a secret, I can finally share some news with you…

I am pleased to announce that the first owner of the new Bugatti Tourbillon will be the wife of the President of Ukraine, Olena Zelenska.

Together with the representatives of Bugatti Automobiles, we organized a private presentation of Bugatti Tourbillon for the Ukrainian delegation.

It took place on June 7, when they visited France. It was 2 weeks before the official announcement of the Bugatti Tourbillon model.

They were the first people to see the new model, and Mrs Zelenska pre-ordered the new Bugatti Tourbillon. 

She will become the owner of the first of 250 cars.

It is a great honor for all members of the Schumacher Group team to work with such an exceptional woman.

We wholeheartedly support Ukraine and are glad that Mrs Zelenska will become a part of the Bugatti family and history”

Authentica Analysis

Authentica offers a comprehensive toolkit of deepfake detection plugins, including several proprietary ones developed by Chess, designed to analyze various physiological aspects of subjects in videos to detect potential manipulation. Each plugin is optimized and weighted for maximum accuracy, depending on the type of video. Although the video was initially identified as a targeted deepfake attack, we conducted further examination using Authentica. After running 13 plugins against the video, our findings confirmed external assessments that the video had indeed been manipulated.

Our analysis generates figures and statistics that highlight evidence of manipulation at increasing level of detail.   For example, the bubble chart in Figure 1 represents the degree to which each plugin classified this video to be real or fake.  Specifically, each plugin (x-axis) returns a probability (y-axis) that a frame is real (0 = fake, 50 ~ suspicion of manipulation, 100 = fake).  The bubble location denotes the average authenticity probability over all the frames, and the bubble size denotes the weighted total number of frames that fell below the 50% manipulation threshold.  

Figure 1:

For example, LipForensics bubble is largest and hovering at a probability of 50%. The LipForensics plugin flagged a considerable number of frames as suspicious, which raises a red flag. Furthermore, taking a look at DFDC1 on the x-axis, while its bubble is fairly smaller than the other red bubbles (meaning less weight/less contribution to the overall classification), its location around a probability of 5% indicates that there is a high likelihood that these frames are fake. 

We can also look at the behavior of the video frames over time. The bar chart in Figure 2 represents the evolution of the manipulation probability for each video frame over time. The color bar key shows the probability range (0 Dark Red - Manipulated, 1 Dark Blue - Real). Some plugins output only one probability over the entire video, which will manifest as one constant color over all the frames. For example, the bar representing DFDC1 is Dark Red during the whole video length, as its probability does not change over video frames.

Figure 2:

Our proprietary CHESS algorithms output a manipulation probability frame-by-frame, thereby allowing us to hone in on which portion of the video was likely manipulated. GenConViT’s Bar alternates between red/yellow/green and blue throughout the video, highlighting sections where manipulations likely occurred and others where no manipulation was detected. 

Deeper Dive

The invoice, intended to bolster the deepfake’s credibility, appears to be a formatted document featuring a Bugatti logo header. It lists the delivery address of the Neuilly-sur-Seine dealership in Paris and specifies that the invoice be sent to “Mrs. Olena Zelenska.” The document includes other details such as the model number, stock number, and relevant dates. It also lists the vehicle’s cost as 4,463,400, though no currency is specified.

Link

Shortly after the incident, Bugatti Paris officially denied the claims on their Instagram account. In their statement, they pointed out several red flags in the supposed invoice: it lacked specific legal details, listed an incorrect price, and featured an outdated visual design. Additionally, there was no mention of an employee named Jacques Bertin, further discrediting the document's authenticity.

The video first appeared on a French-language website, Verite Cachee France (sic), which, upon closer inspection, seemed to be fake or AI-generated. The website, which covers topics related to the war in Ukraine, promotes a pro-Russian viewpoint. BBC’s fact checkers later discovered that a network of Russian websites posing as local publications were behind similar disinformation incidents.

Darren Linvill, a professor at Clemson University’s Media Forensics Hub, told CNN that the deepfake and the Verite Cachee website resemble patterns of Russian narrative laundering that have been ongoing since August 2023. "They typically place some video with a story to tell on YouTube, then layer it onto fake news pages they’ve created, often with allied websites," Linvill explained. "They then amplify the story on social media, starting with real pro-Russian influencers in their network." The only notable difference in this case, he added, was that the video wasn’t posted on YouTube.

A report by Recorded Future identified Verite Cachee as part of the same disinformation network, with intelligence analysts suggesting that the campaign is likely backed by Russian interests. This deepfake attack appears to have been designed to spread distrust among European supporters of Ukraine.

So What?

While the video was quickly debunked, it still reached nearly 18 million viewers within 24 hours. This incident underscores how rapidly disinformation spreads, often going viral on social media before fact-checkers can intervene – an escalating threat in today’s digital landscape. In this case, the narrative likely aimed to amplify themes like Ukrainian corruption and U.S. aid spending, seeking to sway public opinion and reduce support for Ukraine. It also appears to have been timed to discredit Ukrainian leadership ahead of the NATO summit, which began on July 9, 2024.

This deepfake indirectly targeted a public figure by creating a fabricated individual to lend false credibility to the story. It highlights how attackers can exploit both real people and fictional characters to spread disinformation effectively. Hostile foreign states will continue to use these tactics to influence public opinion. As deepfake technology and social engineering techniques advance, incidents like this are likely to increase in 2025 and beyond.

Instead of dedicating your team to search for the latest and greatest deepfake detection tools, we at Chess handle the research, implementation, and optimization of these tools to provide accurate insights for intelligence analysts. We collaborate with open-source intelligence professionals to vet online media and deliver explainable evidence, allowing them to brief government officials and policymakers. Ready to start vetting online media? Get access to the Authentica API here.

Sources:

Gianluca Mezzofiore. “Deepfake Video Targeting Zelensky’s Wife Linked to Russian Disinformation Campaign, CNN Analysis Shows.” CNN, 2 July 2024, www.cnn.com/2024/07/02/europe/deepfake-video-zelensky-wife-intl-latam/index.html.

Myers, Paul , et al. “A Bugatti, a First Lady and the Fake Stories Aimed at Americans.” Www.bbc.com, 2 July 2024, www.bbc.com/news/articles/c72ver6172do.

Reuters Fact Check. “Fact Check: No Evidence Ukraine’s First Lady Bought Bugatti Sportscar in Paris.” Reuters, 13 July 2024, www.reuters.com/fact-check/no-evidence-ukraines-first-lady-bought-bugatti-sportscar-paris-2024-07-13/.