comscore Meta removes Ukrainian President Zelensky deepfake video: What is a deep fake, what’s it used for?

Meta removes Ukrainian President Zelensky deepfake video: What are deepfakes, what are they used for?

Facebook has confirmed that it has removed a deepfake video of Ukrainian President Zelensky asking people to lay down arms. Here we explain what are deepfake video, why have they been banned and what are they used for.


Image: Twitter

Facebook’s parent company Meta has confirmed that it removed a deepfake video of the Ukrainian President Volodymyr Zelensky asking Ukrainians to ‘lay down arms’ in the weeks-old war against Russia. Also Read - Meta Store opens doors to customers: All you need to know

The deepfake video of President Zelensky was first broadcasted by a Ukrainian news site called ‘Ukraine 24’ after an alleged hack. The deepfake video shows President Zelenksy speaking behind a podium declaring that his country had decided to return Donbas to Russia. It also asked the Ukrainian ‘defenders’ to lay down their arms and return to their homes. Also Read - WhatsApp will soon let you set your old messages to disappear at once

In the deepfake video seen by Motherboard’s Vice, Zelensky says, “…I advise you to lay down your arms and return to your families. You should not die in this war. I advise you to live, and I’m going to do the same.” Also Read - Instagram to begin testing NFTs this week in the US: All you need to know

The video doesn’t look doctored but there are telltale signs that point out that the video isn’t what it appears to be. As noted by The Verge, the video shows Zelensky’s head being more pixelated than his surrounding body. Additionally, his voice in the video seems to be much deeper than his real voice.

Confirming the news, Meta’s head of Security Policy Nathaniel Gleicher said, “Earlier today, our teams identified and removed a deepfake video claiming to show President Zelensky issuing a statement he never did. It appeared on a reportedly compromised website and then started showing across the internet.”

“We’ve quickly reviewed and removed this video for violating our policy against misleading manipulated media, and notified our peers at other platforms,” he added in another tweet.

After the video started circulating on social media, President Zelensky posted a video on Instagram debunking the video. “As for the latest childish provocation with advice to lay down arms, I only advise that the troops of the Russian Federation lay down their arms and return home…We are at home and defending Ukraine,” he said in the post.

What are deepfakes?

The term ‘deepfake’ is a combination of two words – deep learning and fake. Now, deep learning is a form of artificial intelligence (AI) that uses large data sets to train itself to solve a problem. In case of deepfakes, scamsters are using AI to create real-looking fake videos by swapping faces in a video.

There are ample examples of deepfakes available on the internet today. Remember the video where Jon Snow apologises for Game of Thrones’ disappointing ending or the video wherein Facebook-founder Mark Zuckerberg brags about having control of billions of people’s stolen data? If you have seen these videos, you have seen a deepfake.

Are deepfakes only videos?

No, not all deepfakes are videos. Deepfake technology can be used for making anything from voice clones to creating photos of people from scratch. Deepfake audios are also used for making deepfake videos more convincing.

How are deepfakes created?

Making a deepfake video isn’t easy. It involves a few steps and a large data set to make a face-swap video that looks real. First, a creator will have to train an AI with hours of video footage of the two people whose faces they want to swap. This will help the AI to understand what the people whose faces the creator aims to swap look like from every angle and under various lighting conditions. Once the AI has been trained, it will find similarities in the two faces and compress the image at the same time. Then a second AI algorithm will be taught to recover the compressed images.

As The Guardian explains it, for making a deepfake video or swapping faces in a video, the creator needs to feed the encoded images of the person whose video needs to be created will be fed in the decoder of the second person, whose video has been recorded for replacement. Since there is a difference between the two videos, the AI will reconstruct the face of the second person with a face that resembles the first person. This process is repeated on every frame of the video and so the end result looks real.

Another technique that creators use to create deepfakes is called generative adversarial network or GAN. In this, two AIs compete with each other. It has two elements – generator and discriminator and the idea is to train the generator AI by making the discriminator AI give its feedback. Overtime both the generator and discriminator improve and in the end, the generator ends up creating realistic images. This technique is used for creating images from scratch. However, they require a bigger data set and take longer to generate images. IEEE Spectrum notes that they are good for synthesising images but not videos.

What are deepfakes used for?

Deepfakes have already been used in the film-making industry. It has been used for inserting Harrison Ford’s young face onto Han Solo’s face in Solo: A Star Wars Story and Princess Leia in Rogue One. But it can also be used for scamming people and mis-represent politicians. In 2018, a Belgian political party released a video of Donald Trump calling on Belgium to withdraw from the Paris climate agreement. Trump never gave that speech – it was a deepfake.

In the same year, Jordan Peele collaborated with Buzzfeed to create a deepfake of the former US President Barack Obama in Peele’s voice to raise awareness about deepfakes.

While these are relatively harmless usages, deepfakes such as the one of the Ukrainian President Zelenksy can topple governments and cause wars.

How to spot deepfakes?

Poor-quality deepfakes are easier to spot. Unnatural blinking, bad lip-synching, problems with hair, skin or face or blurring along the line where the image was patched are some of the ways of spotting a deepfake.

Besides this, tech companies such as Meta, Microsoft, Google and Twitter are also using AI-based algorithms for spotting deepfakes.

For the latest tech news across the world, latest PC and Mobile games, tips & tricks, top-notch gadget reviews of most exciting releases follow BGR India’s Facebook, Twitter, subscribe our YouTube Channel. Also follow us on  Facebook Messenger for latest updates.
  • Published Date: March 17, 2022 2:30 PM IST

new arrivals in india