Meta Platforms has long used artificial intelligence to serve users content and ads it thinks will interest them, but its new chatbot is putting the company's use of the technology back in the spotlight.
The Meta AI chatbot, which is built into the company's platforms (Facebook, Instagram, WhatsApp), can be used to search and answer questions — its promotional page shows it can ask for recipes, generate images and compile a list of popular guitar songs to play.
But some users are wondering how Meta's AI system will use their data, and what they can do to opt out of the new feature.
Where does Meta get the information to train its AI models?
Meta doesn't list all the sources it uses to train its AI models, but the company's privacy center says it uses information that's publicly available online or that it licenses.
The company said it also uses information shared through its products and services, including posts, photos and captions.
What data does Meta not use to train AI models?
Meta's Privacy Center said the company does not use private messages exchanged between users to train its AI systems.
Meta's AI system uses publicly available information collected from the internet and information licensed from other providers, but it doesn't link this data to your account. For example, if you use a public blog post that includes the author's name and contact information, the company says it won't later link that content to your Meta account.
The company also said that the underlying models for Llama 2 and Llama 3 were not trained on user data.
If I don't use Meta's social network, can the company train its AI systems on my data?
It's possible.
“Even if you don't use our products or services or have an account, we may process information about you to develop and improve Meta's AI,” the company's privacy center says.
“This may happen, for example, if you appear in an image shared by someone who uses our products or services, or if you include information about yourself in a post or caption that someone shares on our products or services.”
Can I opt out of Meta using my data?
Meta’s Privacy Center has a form called “Data Subject Rights Regarding Third-Party Information Used for Meta’s AI.”
Users can fill out a form and select “Delete all third-party personal information used to build and improve AI in Meta.”
Meta said it will not automatically fulfill requests submitted using this form, but will instead review the requests to ensure they comply with local laws that may require data to be removed from AI training.
How successful have Canadians been in this process?
Alana Smith, a Toronto-based small business owner in the entertainment industry, filled out a form for Meta to opt out of the company using her data for its AI systems.
She said in an email that the form was “deeply hidden and difficult to find.” She also needed to prove that her data was used to train Meta's AI systems.
“How do we 'prove' that certain data or images are contained in this infinitely large artificial memory that we cannot even see?” she asked.
Smith is unsure if her request was successful because she never received confirmation approving her request.
Mehta spokesman David Troya Alvarez did not respond to questions about how many Canadians have successfully opted out so far.
Besides the opt-out form, is there anything Canadians can do to limit Meta's ability to use their data to train AI?
Media economics professor Brett Callaway said users could consider not using the platform at all, but acknowledged that many people need the company's technology for work and interpersonal communication.
“If I don't want to or can't do that, I try to think twice about what I'm posting,” said Callaway, who has stopped using Facebook but keeps her account inactive.
Rather than satisfying your desire to have information instantly on a platform, he recommends stepping away from your device and deciding whether you really need that information on the platform and whether you're prepared for the company to potentially use it.
Can I disable the Meta AI tool that shows up in Facebook and Instagram searches?
No. Meta spokesperson Troya-Alvarez said in an email to The Canadian Press that Meta AI cannot be disabled.
“Meta AI aims to be a helpful assistant, helping answer users' questions in the search bar,” he said.
“You can't opt out of this experience, but you can still search normally and access different results.”
While you can't turn off the feature, you can mute it by tapping on the blue Meta AI circle in the Facebook or Instagram search bar, then tapping on the blue circle with an “i” inside it. From there, a “Mute” button will appear, allowing you to mute the chatbot.
Does Meta use AI tools to store chat information?
Yes. Meta's privacy center explains that it stores messages you send, the AI's responses, and details about you so it can “keep the conversation going.”
“For example, if you tell our AI that you like Italian food, that detail will be stored,” the company's privacy center said.
“That way, when you ask them later what restaurants they can enjoy in New York, they'll recommend an Italian restaurant.”
How can I see what information Meta has about me in chats with Meta AI tools?
Type /saved-details in the Meta AI chat window or visit the Account Center while logged in to Instagram and Facebook. To download information in WhatsApp, you can also type /download-all-ai-info in a chat with the AI.
How can I delete this information?
Type /reset-ai into each individual chat you've had with Meta AI on Messenger, Instagram, or WhatsApp. Meta says that doing this will delete copies of your messages and the previous context of the conversation, essentially resetting the AI.
“You will still see copies of your chats with the AI, but it will not remember your previous messages,” the company said.
What about other platforms?
Calaway, the University of Toronto's professor, says that even platforms that don't yet use AI models may eventually do so, so it's wise to think about how your information might be used before posting it on a platform.
“Most social media platforms are being encouraged to either develop their own AI tools, if they can afford it, or partner with third-party AI developers,” he says.
“I always assumed that either the platform itself was somehow tied to AI, or there was another company taking all of the publicly available information for AI training. It's just part of the fabric of the internet right now.”
This report by The Canadian Press was first published June 14, 2024.
