How Meta’s AI systems are using your data and why opting out is not so simple

Meta Platforms Inc. has long been using artificial intelligence to help it serve users content and ads it thinks are most likely to interest them, but its new chatbot has thrust the company’s use of the technology back into the spotlight.

The Meta AI chatbot embedded in the company’s platforms — Facebook, Instagram and WhatsApp — can be used for searches and to answer questions. A promotional page shows you can ask it for recipes, to generate images or to put together lists of popular guitar songs to play.

But it’s also prompted some users to wonder how Meta’s AI systems use their data and what, if anything, they can do to opt out of the new feature.

Where does Meta get the information it trains its AI models on?

Meta doesn’t list every source that trains its AI models, but its privacy centre says it makes use of information that is publicly available online or licensed. 

It also says it uses information shared through its products and services, like posts, photos and their captions.

What data does Meta say it doesn’t use to train its AI models?

Meta’s privacy centre says the company does not use private messages exchanged between users for training its AI systems.

While Meta’s AI systems make use of public information collected from the internet or licensed from other providers, it does not link this data to accounts. For example, if it uses a public blog post featuring an author’s name and contact information, the company says it doesn’t later connect that content to the user’s Meta account.

The company also says it did not train its foundational models — Llama 2, Llama 3 — on user data. 

If I don’t use Meta’s social networks, could the company still train its AI systems on my data?

It’s possible. 

“Even if you don’t use our products and services or have an account, we may still process information about you to develop and improve AI at Meta,” the company’s privacy centre says.

“For example, this could happen if you appear anywhere in an image shared on our products or services by someone who does use them or if someone mentions information about you in posts or captions that they share on our products and services.”

Can you opt out of letting Meta use your data?

Meta has a form available in its privacy centre called “Data Subject Rights for Third Party Information Used for AI at Meta.”

Users can fill out the form and select “I want to delete any personal information from third parties used for building and improving AI at Meta.”

Meta says it doesn’t automatically fulfil requests sent using this form but reviews them to see if they are consistent with local laws that might require them to remove your data from their AI training. 

How successful have Canadians been with this process?

Alanna Smith, a Toronto-based small business owner in the entertainment industry, filled out Meta’s form in hopes of getting the company to not use her data for its AI systems.

In an email, she said the form was “deeply hidden and tricky to find.” It also required her to provide proof that her data has been used to train Meta’s AI systems.

“How are we to ‘prove’ that specific data or images are in this infinitely large artificial memory we don’t even have access to review?” she questioned.

Smith is unsure whether her request was successful because she never got a confirmation acknowledging the request. 

David Troya-Alvarez, a Meta spokesperson, did not respond to a question about how many Canadians have successfully opted out so far. 

Other than the opt-out form, is there anything Canadians can do to limit Meta’s ability to use their data for training its AI?

Brett Caraway, a media economics professor, said users can consider not using the platform altogether, but he realizes many people need the company’s technology for their job or interpersonal communication.

“If you’re either unwilling or unable to do that, I tend to think closely about what is it that I’m posting,” says Caraway, who has stopped using Facebook but keeps an inactive account.

Rather than fulfilling the itch to get information on the platform right away, he recommends people step away from their devices and decide whether it really needs to be on the platform and whether they’re prepared for the company to potentially use the information.

Can I disable the Meta AI tool I see in my Facebook and Instagram search?

No. Troya-Alvarez, the Meta spokesperson, said in an email to The Canadian Press that Meta AI cannot be disabled.

“Meta AI aims to be a helpful assistant and is in the search bar to assist with your questions,” he said.

“You can’t disable it from this experience, but you can search how you normally would to engage with a variety of results.”

While you can’t turn the feature off, you can mute it by tapping the blue Meta AI circle in your Facebook or Instagram search bar and then hitting the blue circle with an “i” inside it. From there, you will encounter a “mute” button letting you silence the chatbot.

Is Meta saving information from chats I have with its AI tool?

Yes. Meta’s privacy centre says it saves the messages you send it, the AI’s responsesand details about you so it can “carry on the conversation.”

“For example, if you share with the AI that you like Italian food, it will save that detail,” the company’s privacy centre says. 

“That way if later you ask for restaurants you might enjoy in New York, it should recommend Italian restaurants.”

How can I see what information Meta has on me from chats I had with its Meta AI tool?

Type in /saved-details in your Meta AI chat window or visit your account centre when logged into Instagram and Facebook. To download the information on WhatsApp, you can also type /download-all-ai-info into a chat with your AI. 

How can I delete this information?

Type /reset-ai into individual chats you had with Meta AI in Messenger, Instagram or WhatsApp. By doing this, Meta says you essentially reset the AI because the move deletes its copy of your messages and the previous context of the conversation. 

“You’ll still see your copy of the chat with the AI, but it won’t remember the previous messages,” the company says.

What about other platforms?

Caraway, the University of Toronto professor, says even platforms that aren’t using AI models yet might be inclined to later, so it’s wise to think about how they could use your information before posting on them.

“Most of the social media platforms are incentivized to either pursue their own AI tools, if they can afford it, or to partner with a third-party AI developer,” he says.

“I would always just assume that either the platform itself is somehow tied to AI or there’s another company out there that is ingesting all of the publicly available information for AI training. This is just part of the way the internet is structured now.”