As competing AI models race to develop the most advanced models, everyday internet users are caught in the crossfire. How? By having their personal data—posts, photos, even emails—collected and used to train AI.
This post explains how AI models are training on your personal info – and what you can do to reverse or mitigate it.
How AI Models Collect Your Personal Data from the Internet
Most AI models use absolutely massive amounts of data (think petabytes) as the foundation of their language training. This includes content scraped from the public internet using databases like Common Crawl. Websites, blogs, social media—you name it. But there are more controversial sources too, such as:
- Copyrighted material (e.g., books, artwork)
- Pay-walled material
- Personal and private information
That last category is where things get especially concerning.
For instance, Meta has admitted to scraping both public posts and photos from Facebook and Instagram to train its AI. In the past, Google has used data from emails (sent and received with Gmail), which has even raised concerns that Google Docs could be another future data source.
In short, anything you have publicly posted online—social media uploads, blog posts, forum comments, reviews, etc.—has likely already been scraped. And in some cases, even private communications could be part of the training pipeline.
What You Can Do About It
For data already collected:
Options are unfortunately limited. Depending on where you live, you may be able to request deletion of your data from AI companies. But:
- It’s not guaranteed they will comply.
- This won’t undo the training. Once an AI model is trained on your data, the only way to erase it would be to re-train the model from scratch—which is unlikely.
For future data (or data not yet collected):
You have more control here. Steps include:
- Scrub or privatize your social media. Delete posts or switch accounts to private to help block further scraping.
- Be selective going forward. Only post content you are comfortable being stored, processed, and used to train AI.
- Opt out of crawling. In theory, blocking bots using a
robots.txtfile or privacy settings should, but this may not work. - Switch to private tools. Avoid compromised platforms like Instagram or Gmail. Use a private email service, like ProtonMail, instead.
How Your Interactions with AI Also Become Training Data
Once an AI model has its foundational data, it is often fine-tuned using real-world user input. In simple terms, when you use a chatbot like ChatGPT, the system can learn from how you interact with it.
What You Can Do
- Don’t use AI chatbots. The simplest solution is to avoid them altogether—but that may not be realistic for many people.
- Avoid sharing personal or sensitive info. If you do use AI tools, be careful about the data you provide during interactions.
How AI Uses Your Personal Data: Wrap-Up
As we’ve seen, the privacy landscape is murky when it comes to how AI uses your personal data. If you haven’t already taken steps to limit what you post online or started using more private tools, there’s a good chance some of your information has already been collected.
Still, it’s not all doom and gloom. By taking the steps above, you can reduce future data exposure and avoid contributing more of your personal information to AI training pipelines.
Frequently Asked Questions
Q: Can I stop AI from using my personal data?
A: In many cases, no. Once your data has been collected and used in training, it’s part of the model. Some jurisdictions allow you to request data deletion, but compliance is inconsistent and often doesn’t reverse training.
Q: How do I know if my data has been used?
A: Unfortunately, you usually don’t. Most companies do not disclose specific data sources or individuals whose data was used in training.
Q: Does switching email providers actually help?
A: Yes, especially if you switch to privacy-focused services that don’t use your data for advertising or model training. ProtonMail is one such option.
Q: Can private messages be scraped for training?
A: Technically, yes—especially if those messages are stored on platforms that reserve broad rights to access and use user content (like Gmail or Facebook Messenger).
Q: Is using ChatGPT or other AI tools safe?
A: These tools are useful, but you should avoid sharing personal, sensitive, or identifying information during interactions. Assume anything you write could be stored and analyzed.