Your Images May Be Used To Train AI Tools Without Permission: Can You Opt Out?

Your images may be used to train artificial intelligence, and there's very little you can do to revoke access. Most AI image generators like DALL-E and Stable Diffusion need to be trained on large datasets of images that are scraped from the internet.

The downside? It is most likely for your images to have made into the AI database. Even if you never agreed to let the AI use your images, there aren't many options in  case you want to opt out. Remember Facebook 10 years ago? Yep, there's where we're at with AI right now.

iStock

The dangers of AI datasets

The problem with these huge datasets is that they're designed to scrape images that they do not own. These images may not be classified correctly and copyright holders and people in the images may not have given their permission to train AI tools.

One of these datasets is the LAION dataset that was used to train Google's Imagen and Stable Diffusion. In one instance, a Discord user highlighted how their image taken by a doctor for documentation somehow made it into the dataset.

Also read: A Terrifying Woman Named 'Loab' Keeps Appearing In AI Generated Images

iStock

A LAION dataset engineer, Romain Beaumont, told Vice that to take down these images, the best route is to ask the host website to stop hosting it. But finding where all your image may be hosted might be tricky. But Beaumont said that "that's what the dataset is," adding that one can get full list of URLs by downloading the datasets.

Earlier, the LAION-5B dataset, that has 5 billion images as the name suggests, included photoshopped celebrity porn, stolen nonconsensual porn, and even ISIS beheadings. In addition, there were images of art, and of people who may not be aware their images are used to train AI.

Can you see if you've been used?

AI tools like DALL-E and Midjourney offer no way for users to find out if their images have been used, for they're not open-source. LAION, that claims it is a "truly open AI" but even then, there's no clarity on how images scraped by a giant AI tool may be removed.

iStock

Also read: China's New AI Image Generator Censors Words Like 'Revolution,' 'Tiananmen Square'

What's worse is that not many people would ever find out that their images have been into an AI's training process. If you want to check out the same, you may use tools like "Have I Been Trained." The tool, created by artist Holly Herndon is a simple way to check if your pics are being used in the LAOIN-5B dataset.

What do you think about your images making their way into an AI dataset without your knowledge? Let us know in the comments below. For more in the world of technology and science, keep reading Indiatimes.com.

References

Xiang, C. (2022, September 26). AI Is Probably Using Your Images and It’s Not Easy to Opt Out. Retrieved September 27, 2022, from https://www.vice.com/en/article/3ad58k/ai-is-probably-using-your-images-and-its-not-easy-to-opt-out