Artificial intelligence

Thanks to AI, it’s probably time to take your photos off the internet

Enlarge / It’s John. He does not exist. But the AI ​​can easily put a picture of him in any situation we want. And the same process can apply to real people with just a few real photos pulled from social media.

Benj Edwards / Ars Technica

If you are one of Billions people who have been posting photos of themselves on social media over the past decade, maybe it’s time to rethink that behavior. New AI image generation technology allows anyone to save a handful of photos (or video images) of you, then train the AI ​​to create realistic fake photos that show you doing embarrassing or illegal things. Not everyone may be at risk, but everyone should know.

Photographs have always been tampered with, first in darkrooms with scissors and glue, then through Adobe Photoshop through pixels. But it took a lot of skill to pull it off convincingly. Today, creating compelling photorealistic counterfeits has become almost trivial.

Once an AI model learns to render someone, their image becomes a software toy. The AI ​​can create images of them in infinite quantities. And the AI ​​model can be shared, allowing other people to create images of that person as well.

John: A social media case study

When we started writing this article, we asked a brave volunteer if we could use his social media images to try and train an AI model to create fakes. They accepted, but the results were too convincing and the reputational risk proved too great. Instead, we used AI to create a set of seven simulated social media photos of a fictional person we’ll call “John.” This way we can show you the results safely. For now, let’s assume John is a real guy. The result is exactly the same, as you will see below.

In our fictional storyline, “John” is an elementary school teacher. Like many of us, over the past 12 years, John has been posting photos of himself on Facebook at work, relaxing at home, or on the go.

These innocuous images, inspired by social media,
Enlarge / These innocuous images of “John” on social media were used as training data which our AI used to place him in more compromising positions.


Using just these seven images, someone could train the AI ​​to generate images that make it look like John has a secret life. For example, he might like to take nude selfies in his class. At night, John can go to bars dressed as a clown. On weekends, he could be part of an extremist paramilitary group. And maybe he served time in prison for an illegal drug charge, but hid it from his employer.

We used an AI image generator called Steady broadcast (version 1.5) and a technique called Dream Cabin to teach the AI ​​how to create images of John in any style. Although our John is not real, someone could replicate similar results with five or more images of any person. They can be pulled from a social media account or even taken as stills from a video.

Leave a Reply