Deepfake video shows Zelensky ordering Ukrainian troops to surrender

Deepfake video shows Zelensky ordering Ukrainian troops to surrender
	MANIPULATED MEDIA
	"A video released in March 2022 shows Ukrainian President Volodymyr Zelensky telling Ukrainian soldiers to lay down their arms and surrender to Russian forces."
Domain	Technology & Media
Tactic	Manipulation; Imposture; Propaganda Technique;
Era	Current
First seen	2022
Origin	Posted on a hacked Ukrainian television website and spread on social media and Russian-language channels on March 16, 2022
Last reviewed	2026-10-26
	Further reading: books on Technology & Media

The deepfake video of Volodymyr Zelensky ordering Ukrainian troops to surrender appeared on March 16, 2022, three weeks after Russia's full-scale invasion of Ukraine. It showed what appeared to be the Ukrainian president standing at a podium, telling soldiers to put down their weapons and return to their families. The video was fabricated using artificial intelligence. It was quickly identified as fake, removed by major platforms and rebutted by Zelensky himself within hours. It is widely regarded as the first significant use of a deepfake in an active war, and an early test of whether such fakes could be used to undermine a country's will to fight.

The claim

In the video, a figure resembling Zelensky, speaking in Ukrainian, said that he had decided to return Donbas to Russia and urged Ukrainian soldiers to stop fighting. The message directly contradicted Zelensky's actual public stance, which was one of determined resistance. Had it been believed, even briefly, by soldiers or civilians, it could have caused confusion at a critical moment in the war.

Origin

On March 16, 2022, hackers compromised the website and live news ticker of Ukraine 24, a Ukrainian television channel, and posted a message claiming that Zelensky had surrendered. The deepfake video was posted online and spread on social media platforms, including Facebook, YouTube, Twitter, Telegram and VKontakte, a Russian social network.

Ukrainian officials had anticipated such an attack. On March 2, 2022, Ukraine's Center for Strategic Communications, part of the government, warned that Russia might release a deepfake video of Zelensky announcing surrender, and urged citizens not to believe it.

What the evidence shows

The video was quickly identified as fake for several reasons:

Visual flaws. Zelensky's head appeared too large for his body and was lit differently from the rest of the image. His face moved unnaturally, and his body was largely static.
Audio flaws. The voice did not sound fully like Zelensky's and had a different accent and tone.
Contradiction with known facts. Zelensky was in Kyiv and had been releasing regular, defiant video messages to the nation.
Rapid denial. Zelensky posted a video on his official Instagram account in which he dismissed the fake as a childish provocation and said that Ukrainians would not lay down their arms.

Meta, the parent company of Facebook, said it had quickly identified and removed the video for violating its policy against misleading manipulated media. YouTube and other platforms also removed it. The origin of the video was not publicly confirmed, but Ukrainian officials attributed it to Russia, and the hack of a Ukrainian broadcaster fit a pattern of Russian information operations during the war.

Significance

Although the Zelensky deepfake failed, it was important for several reasons:

First wartime deepfake. It was one of the first known uses of a deepfake to impersonate a head of state during an armed conflict, a scenario that researchers had warned about for years.
Prebunking worked. The government's warning two weeks earlier meant that many Ukrainians were primed to recognize a fake. Researchers have pointed to the episode as an example of the value of warning people in advance about specific kinds of manipulation.
Speed of response. Zelensky's quick denial and the platforms' rapid removal limited the video's impact.
Quality will improve. The video was crude by later standards. As AI tools have advanced, convincing fakes have become easier to make, and later incidents in elections and fraud cases have been harder to detect.

The episode also illustrates the so-called liar's dividend, a term coined by legal scholars Robert Chesney and Danielle Citron. As people become aware that deepfakes exist, it becomes easier for anyone to dismiss genuine evidence as fake. Information warfare can work not only by making people believe false things but also by making them doubt true ones.

Why it persists

The Zelensky deepfake remains a reference point in discussions of AI and information warfare. It is cited as both a warning and a reassurance: a warning that hostile actors are willing to use such techniques, and a reassurance that preparation, fast response and public skepticism can blunt them. It is also frequently referenced when similar fakes appear, such as fabricated audio and video of political leaders in later elections around the world.

Lessons for readers

Be especially skeptical of dramatic statements from leaders that contradict everything they have said before.
Check official channels. Leaders and governments usually confirm or deny major announcements quickly on their verified accounts.
Look for visual and audio inconsistencies, such as mismatched lighting, unnatural facial movement or an unfamiliar voice.
Prebunking helps. Knowing in advance that a particular kind of fake may appear makes it easier to recognize.
Remember the liar's dividend: the existence of deepfakes can also be used to cast doubt on real evidence.

Sources

Bobby Allyn, "Deepfake video of Zelenskyy could be 'tip of the iceberg' in info war, experts warn," NPR, March 16, 2022.
Nathaniel Gleicher, Meta head of security policy, statement on removal of the Zelensky deepfake, March 16, 2022.
Center for Strategic Communications and Information Security (Ukraine), warning about planned deepfakes, March 2, 2022.
Robert Chesney and Danielle Keats Citron, "Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security," California Law Review 107 (2019).