What Are Deepfakes?

A deepfake is a form of media, such as a photo or video, generated by Artificial Intelligence (AI) to depict real or non-existent people performing actions they never did. AI manipulates a picture, video or voice recording to analyze a person’s characteristics and then blends those characteristics with existing footage using unique algorithms. Deepfake technology can be used to show a person saying something they have never said or doing something they have never done by placing their face, body or voice onto another person.

Continue reading to learn why deepfakes are so dangerous, how they are created, whether they are legal and how to spot them online.

The dangers of deepfake technology

Although deepfake technology is not new, it’s getting better at making media look more believable due to the rise of AI tools. There are several ways deepfake technology can be used with malicious intent, such as impersonation, blackmail and fraud. Someone can create a deepfake to spread misinformation or damage someone’s reputation if an AI-generated photo or video places someone in an inappropriate or illegal situation. The implications of a fake photo, video or voice recording can cause mass confusion, spread lies and ruin someone’s public image.

How are deepfakes created?

You may be wondering how such realistic images, videos and audio recordings are made. Generating deepfakes requires an extensive amount of data, primarily about the person being targeted. Once enough photos and videos have been collected, AI analyzes everything about that person, including their facial expressions, body movements and unique features. At the same time, AI will get to know the data from the person’s image or voice that is being added. By blending the target person’s details, AI can create a seemingly natural-looking image, video or voice recording. From there, slight movements and modifications to a person’s shadows, lip movements, blinking and more can be made to make the photo or video appear as natural as possible.

Are deepfakes legal?

Despite the vast harm that could be done to victims of this technology, deepfakes are not entirely illegal. However, some states have begun to outlaw deepfakes depending on their content and intent. Even though several states are taking action against this technology, there is no federal law currently addressing the creation and distribution of deepfakes. The DEEP FAKES Accountability Act was first introduced to Congress in 2019; if passed, it would create new criminal offenses related to deepfakes and provide victims the ability to seek legal action.

How to spot deepfakes

Since deepfakes are becoming more realistic by the day, it can be challenging to detect whether or not a piece of media is real. However, there are several things you can look for to spot a deepfake and prevent yourself from being deceived or misled by false images or videos.

Lack of natural blinking

In a deepfake video, pay close attention to how natural a person’s blinking appears. If the person’s blinking is inconsistent or their eye contact does not match the direction of their face, this is a sign that a person’s face may have been superimposed onto another person’s body as a deepfake. Another indicator of unnatural blinking in a deepfake video can actually be seen by looking at the subject’s eyes, as a deepfake will generally have pixelated eyes if examined closely.

Mismatched audio

When listening to a video or voice recording, you can tell if it has been manipulated as a deep fake by noting any audio distortions. You may hear strange echoes or background noise inconsistent with the video’s intended background. For example, if a celebrity is speaking and the audio has downtown traffic noises in the background, but the video shows them in a barren cornfield, the video is most likely a deep fake.

Odd facial positioning and expressions

An obvious sign of deepfakes is how a person’s face moves in a video. Human beings have natural body movements and facial expressions, but AI struggles to generate an accurate depiction of a person running or laughing, even with plenty of references. If you see a video of a person chasing after someone else and their body does not move as it should, this could be a deepfake. Also, take a close look at the shadows of a person in a deepfake photo or video, as AI often fails to account for natural lighting.

Unnatural textures

Deepfakes typically use photo or video-enhancing software to smooth a person’s skin so much that you can hardly see any wrinkles, pores or blemishes. People have also noticed that many AI technologies used to create deepfakes cannot distinguish individual teeth, so if a person is smiling and you cannot identify separate teeth, this could mean the image is a deepfake. Similarly, AI also tends to blur individual body parts, such as fingers on a hand. Overly blurred or smoothed images could be deepfakes since AI does not always understand the context in which it is supposed to generate a person.

Don’t fall for deepfakes

Deepfakes can be extremely difficult to distinguish from real images, videos and audio recordings as AI continues to improve its ability to generate more accurate content. However, it’s important to be aware of the common signs of deepfakes to avoid believing lies or spreading misinformation.