Your own old posts and photos are training AI, right now

AI companies need a huge amount of text and images to train their models on, and pretty much all of it is just scraped straight off the public internet. Yes, that literally means posts, comments, profile pictures, and all the other stuff you posted online years ago and haven't thought about since, and in most cases nobody ever asked you about it, or even told you that this was happening.

People have been worried about this for a long time, but it is already happening at massive scale, there has been court cases about it; the numbers are insane.

Your own face

The scariest example is Clearview AI, a facial recognition company which built its whole product by scraping photos of people from social media, news sites and even services like Venmo. By the time they ended up in court, they already had put together a database of more than sixty billion facial images. Over sixty billion images taken from public accounts without asking a single person, and then sold straight to the police and other shady private companies.

In March of 2025, a US court approved a settlement in the case, which was brought under the Illinois' biometric privacy law. The settlement itself was pretty unusual, because instead of getting paid a cash settlement, the people were literally given a shared 23% equity stake in the company, so basically a tiny (per person), share in the same company that had been built using their faces in the first place. This share, on an individual level, would've amounted to pocket change given the amount of people actually affected.

Your social media posts

It's not just your face, and it's also not just random private companies, that you've never heard of, doing it. Last year, in 2025, the biggest platforms openly started using people's own public posts and comments to train their own AIs. Meta, for example, told their users it would be training its AI models on public content from its platforms, Instagram and Facebook, and acted as if it were something they are allowed to without asking. If you have posted things publicly over the years, the default assumption that you agreed to it would have already worked against you, and it is mostly too late now. Unless you want to delete your account and stop providing them more data, of course.

A lot of the raw data is just shared around in the open too. Datasets such as LAION) put together billions of scraped images and text so that anyone can use them to train their own AIs. In 2024, and 2025 the courts have started ruling on whether that should have even been allowed in the first place. The law genuinely caught up yet, as usual, and by now it is mostly too late.

You can't unscrape it, but you can cut off the source

Once your data has been scraped and used to train a model, it's basically irreversible at that point. There's really no way to go into a trained AI and take your face or your posts back out of it. That is a big part of why these companies want to scrape so much so quickly, they are trying to collect as much as they can before any stronger regulation appears.

The scraping doesn't just happen one time and then it's done, it is ongoing. These companies keep coming back for more and more data, and every new model that gets released gets trained on whatever information is public at the time. So, whatever is still sitting on your old accounts and has not been trained on yet, it's only a matter of time before it does. That public Facebook account with pictures of you as a kid, that you haven't logged into since 2015 is still out there, and it will one day be scraped for the another round of training.

Deleting those old accounts means there's nothing left on them to scrape. It won't undo anything that has already been taken unfortunately, but it will stop the old forgotten parts of your digital footprint and social media presence from getting thrown into the next AI, and so on and so forth. So it is genuinely worth looking up how to delete the old accounts you don't use anymore, like a profile you abandoned, an old photo hosting site, or some forum, before your own creations and your own photos get used without your consent, again.

We built removi for this. So, rather than you having to manually track down dozens or hundreds of accounts by hand, it automatically finds the old ones that are linked to your email and deletes them for you, so there is a lot less of you out there for them to take.

Sources: the Clearview AI settlement (60 billion scraped images, 23% equity remedy, approved March 2025), coverage of platforms training AI on public posts, and the LAION court rulings) on scraped training datasets.