Soon, Facebook will use the things you post to train its AI, and it’s not easy to stop this.

Meta likely made the process arduous for a reason


It won’t be long before Facebook is the next site to use posts and pictures from its users to train its AI models. People can choose not to participate, but Meta seems to have purposely made the process hard and time-consuming. It’s also likely to fail.

On June 26, 2024, Facebook’s privacy rules will change. From then on, public user information, not private conversations, will be used to train and improve its AI services.

Meta gives you a way to say no to data scraping, but it’s not as easy as checking a box.

Visit this help page and click on the spot that says, “I want to object to or limit the processing of my personal information from third parties utilized for building and enhancing AI at Meta.” If you still enjoy opt-out, you can do so.

After that, you just need to type in your email address, name, last name, and the country where you live. The form asks if Meta’s AI models know anything about you. This includes giving any information about the questions you asked that led to your personal information showing up in a Meta model, feature, or experience answer. Meta also wants you to show an image that shows your private information in an AI model’s answer.

In the last part, there is an extra context box that asks you to give a short description of your problem and what you want.

A user named Tantacrul on X says that Meta is also telling users about the new AI tools. Following a click on this and then a “right to object” link will lead you to a similar form with a part for “how this processing impacts you.” Meta also asks users to enter an OTP that was sent to their email address. The company says this is for security reasons, but the fact that the entry won’t be sent until the code is entered makes it look like it’s just another way to get people not to fill out the form.

There is a note from Meta that says, “If your objection is honored, it will be applied going forward.”

Tantacrul writes that the company did listen to his argument, but many of his friends and Reddit users who also did so got an error message when they tried to send the form.

It’s becoming common to use scraped user material to teach AI models new things. This month, OpenAI teamed up with Reddit to add user posts to ChatGPT.

Comments are closed.