A viral video recently surfaced on X (formerly Twitter), featuring what appeared to be popular actress Rashmika Mandanna in sportswear entering an elevator. However, it was soon revealed that the video was a deepfake. The original clip actually featured Instagram influencer Zara Patel, and using AI-powered applications, miscreants morphed Mandanna’s face onto Patel’s.
What Is A Deepfake?
- Deepfakes use AI to generate completely new video or audio, with the end goal of portraying something that didn’t actually occur in reality.
- The term “deepfake” comes from the underlying technology — deep learning algorithms — which teach themselves to solve problems with large sets of data and can be used to create fake content of real people.
- A deepfake would be footage that is generated by a computer that has been trained through countless existing images
- Deepfakes aren’t just any fake or misleading images.
- The AI-generated fake scenes of Donald Trump being arrested that circulated shortly before his indictment, are AI-generated, but they’re not deepfakes.
- What separates a deepfake is the element of human input.
- When it comes to deepfakes, the user only gets to decide at the very end of the generation process
- if what was created is what they want or not; outside of tailoring training data and saying “yes” or “no” to what the computer generates after the fact, they don’t have any say in how the computer chooses to make it
How Are Deepfakes Created?
- There are several methods for creating deepfakes, but the most common relies on the use of deep neural networks that employ a face-swapping technique. You first need a target video to use as the basis of the deepfake and then a collection of video clips of the person you want to insert in the target.
- The videos can be completely unrelated; the target might be a clip from a Hollywood movie, for example, and the videos of the person you want to insert in the film might be random clips downloaded from YouTube.
- The program guesses what a person looks like from multiple angles and conditions, then maps that person onto the other person in the target video by finding common features.
- Another type of machine learning is added to the mix, known as Generative Adversarial Networks (GANs), which detects and improves any flaws in the deepfake within multiple rounds, making it harder for deepfake detectors to decode them.
- Though the process is complex, the software is rather accessible. Several apps make generating deepfakes easy even for beginners — such as the Chinese app Zao, DeepFace Lab, FakeApp, and Face Swap — and a large amount of deepfake softwares can be found on GitHub, an open source development community
How To Detect A Deepfake?
There are a handful of indicators that give away deepfakes:
- Do details seem blurry or obscure? Look for problems with skin or hair, or faces that seem to be blurrier than the environment in which they’re positioned.
- Does the lighting look unnatural? Often, deepfake algorithms will retain the lighting of the clips that were used as models for the fake video, which is a poor match for the lighting in the target video.
- Do the words or sounds not match up with the visuals? The audio might not appear to match the person, especially if the video was faked but the original audio was not as carefully manipulated.
- Does the source seem reliable? One technique that journalists and researchers often use to check the true source of an image, which you can do right now, is reverse image searching.
Government Rules to Curb Deepfakes
- IT Act, 2000 and IT Rules, 2021: Both the IT Act and IT Rules have clear instructions which place the onus on social media intermediaries to ensure such deep-fake videos or photos are taken down as soon as possible. In case of failure, there are provisions for imprisonment of up to three years of fine of Rs 1 lakh.
- Section 66D of IT Act: Section 66D of the IT Act, 2000 states that anyone who cheats by personating using a communication device or computer resource can be punished with Imprisonment of up to three years and a fine of up to one lakh rupees.
- Rule 3(1)(b)(vii): This Rules mandates that social media intermediaries must ensure that the users of their platform do not host any content which impersonates another person.
- Rule 3(2)(b): It requires such content to be taken down within 24 hours of receipt of a complaint against such content.
- Section 500 of the Indian Penal Code (1860) offers punishment for defamation, which can be applied in cases involving deepfakes.
Global Regulation to Curb Deepfakes
European Union: The European Union’s Code of Practice on Disinformation requires tech companies to counter deep fakes and fake accounts within six months of signing up to the Code. If found non-compliant, tech companies can face fines up to 6% of their annual global turnover
United States: The U.S. introduced the bipartisan Deepfake Task Force Act to assist the Department of Homeland Security in countering deepfake technology.
China: China introduced comprehensive regulation on deep synthesis, effective from January 2023. Aimed at curbing disinformation, the regulation requires clear labelling and traceability of deep synthesis content.
The Role of Generative Artificial Intelligence (GAI) in Deepfakes
- GAI is a rapidly growing branch of AI that focuses on generating new content (such as images, audio, text, etc.) based on patterns and rules learned from data.
- The rise of GAI can be attributed to the development of advanced generative models, such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs).
- These models are trained on large amounts of data and are able to generate new outputs that are similar to the training data. For example, a GAN trained on images of faces can generate new, synthetic images of faces that look realistic.
- Arguably, because machine learning and deep learning are inherently focused on generative processes, they can be considered types of GAI, too.
To Download Monthly Current Affairs PDF Click here
Click here to get a free demo
Everything About CLAT 2026
