Online issues and worries: Generative AI

A deepfake uses AI to change videos, images or audio of real people to make it seem like they are doing or saying something that they aren't.

A lot of the time, the featured people (or subjects) are public figures like celebrities or politicians. But sometimes, a deepfake might feature an everyday person like someone from school. It can be really scary to see a deepfake of yourself online.

Deepfakes can only be created if a human has shared real pictures of the person with the AI. For public figures, there are a lot of pictures and videos online that someone can use. For other people, someone might get images and videos from social media to create deepfakes.

Why people create deepfakes

People might create deepfakes for different reasons. While some reasons might be innocent, others could be really negative. Either way, the deepfake is taking away the subject’s control of how they present themselves.

For fun

Some people like to play around with AI tools to see what they can create. However, they should use their own likeness to do this instead of someone else’s, especially if they can't get permission to use the subject's image.

To bully

Some people might create deepfakes to bully someone. To do this, they might create a deepfake where the subject is doing or saying something embarrassing. The bully then shares the deepfake with others (maybe around school or online) as if it’s real to target the victim.

To mislead

Some people might create deepfakes so that the people they share it with believe something untrue. This is called disinformation.

For example, they might make it seem like a politician is saying something shocking or hateful to try and get people to mistrust that politician. If lots of people start sharing this online as something that's true (misinformation), it can have huge impacts on people’s beliefs and might make them consider voting for someone else.

To cause anger or fear

Some people might find it funny to make people scared or angry. So, they create a deepfake that shows someone doing or saying something wrong. People who make others angry or scared online for fun are called trolls.

Extortion

Extort means to threaten someone unless something is done. Usually that ‘something’ is paying a certain amount of money.

If a person is trying to extort someone, they might create a deepfake of that person in an embarrassing or humiliating situation and threaten to release it unless they get what they want.

A common type of deepfake used to extort someone is a nude deepfake. This is where the subject appears to be naked or partially naked when they never actually were.

To scam someone

Voice-cloning is a type of deepfake that uses a subject’s voice. It’s used in some scams to trick people into paying them money.

For example, a mum might get a call from her child who is in trouble and needs money right away. The mum panics and sends the money to help her child. What she doesn’t know though is that her ‘child’ is actually a scammer using an audio deepfake to trick her.

Scammers like this prey on the panic someone might feel to hear their family member or friend is in trouble.

How to tell if something is a deepfake

With AI getting better at creating realistic deepfakes, it’s hard to tell real video and images from fake ones.

There are some common signs to look for to check if something is a deepfake, including:

• the voice doesn’t sound right, for example it sounds robotic or not quite like what the person normally sounds like
• parts of the video seem to move in a weird way, for example the person’s mouth doesn’t always line up with what their voice is saying
• the person's hands, face or other body parts look unnatural

While these things can help you spot some deepfakes, not all deepfakes will have these signs, so it's more important to fact-check and think critically.

Even if a video or image looks real, you shouldn’t share it without checking it first. If you do, you might be spreading misinformation that could cause fear or anger. Plus, those deepfakes spreading harmful lies can damage the subject’s reputation along with people’s trust in them.

Tips to fact-check videos and images

Search the story on a fact-checking website

Website like FullFact.org or Snopes.com review popular stories on the internet and tell you if they’re true or false.

Find out what other people think

Can you find the story somewhere else? If you see a video on a website where people can comment, see what they are saying. Are some people saying it’s fake? If you do this, make sure you check other sources too.

Try a reverse image search on Google

1. Go to Google.co.uk.
2. In the search bar, look for the icon next to the microphone that looks like a camera. Click the camera icon.
3. If you have the image saved on your computer or device, upload it here. If you have the direct image URL (it should end in .JPG, .PNG or something similar), you can paste it instead. Once you add the image, you’ll see results of similar images.
4. Above your image, click ‘Find image source’. You’ll then see different websites that have the image.
5. Now, it’s time to think critically. Are these websites trustworthy? Do the websites support the story you saw or say that it’s fake?

If you’ve tried to fact-check the videos or images you see but are still confused, that’s okay. Talk with a trusted adult such as a parent, carer, teacher or grandparent to see if you can figure it out together.