Can you really believe what you see online anymore? Even faces, voices or words of famous people, influencers—or even your close friends—can now be fake. These are called AI Deepfakes.
They can disguise themselves as another person, whether through face, voice or speech by using Artificial Intelligence to create fake videos, sound or photos that really look real. They can lead to serious consequences like fraud & scams, political manipulation, misinformation or online harassment.
So how do we knowingly and protect ourselves from being fooled?
Want to know how AI deepfakes work and how dangerous they are? Read on below.
What’s AI Deepfakes?
AI Deepfakes are the fake media created with Deep Learning AI using a specific technique called GANs or Generative Adversarial Networks to create very realistic fake media. They can imitate faces, voices or human’s behaviors to create fake media for various purposes.
Or simply call “Using AI disguising other people through videos, voices or images.”
History and Evolution
Beginning of AI Deepfakes
Deepfakes technology was well-known in 2017 by a Reddit user called “deepfakes”. Starting by posting fake videos using celebrities’ faces created by Machine Learning technology and AI to make the computers know and can imitate people’s faces.
These videos became viral online and introduce us to new word “Deepfake” which is from Deep Learning + fake
Technology Evolution
- In 2014 lan Goodfellow introduced GANs or Generative Adversarial Networks which is the crucial fundamental of Deepfake.
- In 2017 Deepfake videos were published on Reddit with a Face-Swap technique by AI.
- In 2018-2019 Deepfake tools were developed for user-friendly applications such as DeepFaceLab and FaceSwap.
- 2020 onwards Deepfakes were mostly used for crime, politics, advertising or entertainment industry.
How do AI Deepfakes work?
- Gathering original information
AI will gather a lot of information like videos, photos or voices of people who want to imitate it to learn faces, shapes, behaviors and voices.
- Practicing AI by Deep Learning
Using a GANs or Generative Adversarial Networks technique which consist of 2 parts
- Generator : Create fake information that are images and sounds
- Discriminator : Check whether what is real or fake
- Create fake media
AI will learn and remember people’s faces, voices or behaviors to paste in new contexts.
- Put one person’s face in another person’s video
- Imitating person’s voice
- Changing face, mouth and voice realistically
- Rendered as Deepfake media
The system processes all to be realistic videos or sounds that are difficult for ordinary people to distinguish.
Want to know why this technology is dangerous? Continue to “Common Uses of AI Deepfake” below.
Common uses of AI Deepfakes
Even there are many criticisms about the ethics, safety and appropriateness of using AI Deepfakes. They use in many different purposes both positive and negative ways.
Positive Usage
- Film Industry
- Using Deepfake for making characters in the films age differently like getting younger or older.
- Re-create dead actors to be in the film such as Paul Walker in Fast & Furious.
- Lucasfilm use Deepfake to create younger face of Luke Skywalker in The Mandalorian series.
- Game
- Creating realistic game characters
- Marketing and Advertising
- Use celebrity faces without actually filming.
- Pepsi Ad: Using Deepfake to create Salman Khan in two ages
Negative Usage
- Fake News
- Distorting information to influence public opinion.
- Create videos of the politicians saying something they never really said.
- Fraud & Scram
- Imitating famous people or CEO’s voices to order money transfer.
- Create fake videos that disguise themselves as the acquaintances to deceive people into transferring money.
- Online Haressment
- Using ordinary people’s faces in inappropriate videos.
- Causing defamation or damage to reputation.
- Privacy Violation
- Creating fake videos using images of ordinary people to mislead or harm others.
The Dark Side of AI Deepfakes
Nowadays, we criticize AI Deepfakes a lot on social media in terms of safety, dangerous, or inappropriate usage which mostly used the negative ways. It causes many negative impacts in our society, like defamation, damaging reputation or harassment.
How Big Tech Responds to Deepfakes
Due to the negative impact and problems happening from AI Deepfakes, many countries have sticker measures to control or some countries enforce laws to ban. The major and big platforms like Google, Facebook and X starting to have control measures Deepfakes that damage or harm users.
Deepfake Cases Around the World
There are many cases happening in various fields whether politics, fraud & scam and online harassment. And, the main cause is AI Deepfakes. Let me show you the danger of AI Deepfakes through many cases.
Politics Cases
Deepfake technology widely use for political purposes by creating fake videos and images that put politicians’ faces in another video or imitating voices to spread fake news, create division, political manipulation and serious consequences that damage politics and international relations.
- Fake video pretends to be Ukrainian President Volodymyr Zelensky ordering surrender
In 2022, Ukrainian TV station was hacked and it broadcasted Deepfake video of President Volodymyr Zelensky. In this video, he ordered Ukrainian soldiers to lay down their arms and surrender to the Russian army.
- Edited Deepfake video of Nancy Pelosi
A video of Nancy Pelosi, a United States house of representatives member publishing on Facebook. It was edited to pretend that she was drunk and slurred her speech abnormally.
Fraud & Scam Cases
Due to AI Deepfakes can create realistic fake media that are hard for ordinary people to distinguish, scammers then use this to create fake media of famous people for scam people. Causing a lot of people losing their money and affecting their mental health.
- Deepfake video of Elon Musk
Fake video of Elon Musk created by AI spread quickly and widely online. This video is really realistic by imitating his voice and image. Causing Steve Beauchamp, a 82 years old retired man, to lose over $690,000 invested in this scam.
- Deepfake disguise as a CFO and deceived employees into transferring more than 25 million US dollars.
Arup, a famous engineering company in the UK, was deceived by Deepfake in an online meeting. The Deepfake disguised as a CFO (Chief Financial Officer) and other employees and made one employee lose more than 25 million US dollars.
Online Harassment Cases
98% of Deepfake videos are pornography and the majority of victims are women. AI Deepfakes mostly used to harm and harassment people online. Causing them to feel insecure, ashamed and affect mental health.
- Nth Room, a online chat room on Telegram, spread and sold pornography images of South Korea women created from AI Deepfakes.
In 2018-2020, Nth Room was the platform that sold and spread pornography images created by AI Deepfakes to imitate faces of women in South Korea. Moreover, there were more than 103 female victims, including 26 minors, who were sexually abused and forced to film themselves.
- Pornography Deepfake of Female students in India.
Profile pictures of female students on Instagram became Deepfake porn. There were many victims who felt stress, depression and didn’t dare to go to school.
How to Detect and Prevent
- Notice the abnormal of images and videos
Try to look carefully and notice lip movements, skin tone or strange lighting, unnatural expression and voice.
- Use tools to detect
There are many platforms to detect Deepfakes such as Microsoft Video Authenticator, Deepware Scanner and Sensity.ai.
- Verify sources of videos
Check the sources whether they are reliable sources or not.
- Media Literacy
Teach and educate people especially children and the elderly to have media literacy, learn about Deepfake technology.
What’s Next?: Future of AI Deepfakes
AI Deepfakes in the future can be both the new opportunity and more dangerous harm. On one hand, Deepfakes technology will improve to be more realistic, user-friendly and create various types of fake media for using in many industries like entertainment, marketing and advertising. However, the most common usage is cyber crime. AI Deepfakes are like a double-edged sword of the AI era. Therefore, we have to prepare for coping with change and development.
Share this article to make people around you aware of the online dangers.
