scorecard
  1. Home
  2. artificial intelligence
  3. news
  4. Your old Facebook and Instagram posts were probably used to train Meta's AI. You can't opt-out as an American, but you can do this instead.

Your old Facebook and Instagram posts were probably used to train Meta's AI. You can't opt-out as an American, but you can do this instead.

Sarah Jackson   

Your old Facebook and Instagram posts were probably used to train Meta's AI. You can't opt-out as an American, but you can do this instead.
  • Meta recently admitted to scraping public posts and photos outside of the EU dating back to 2007 to train its AI models.
  • Users in the European Union can opt out of having their data used to train Meta's AI.

That ill-advised mirror selfie you posted back in college? It's already been ingested by Meta's AI to make its systems more capable — unless you had the foresight not to select "public" when you posted it or set your profile to private.

We now know that Meta has scraped the public Facebook and Instagram data, including things like posts and photos, of adults dating back as far as 2007 in order to train its AI models.

The company's global privacy director, Melinda Claybaugh, made the admission when pressed during an inquiry into the matter in Australia.

While users in the European Union can opt out of having their data used to train Meta's AI models, such an option doesn't exist in the US or Australia.

Making sure your audience isn't set to "public" should keep any of your new posts and photos from being used for the company's AI training. The section on generative AI from Meta's privacy center says, "We use public posts and comments on Facebook and Instagram to train generative AI models for these features and for the open-source community. We don't use posts or comments with an audience other than Public for these purposes."

A blog post from September 2023, when the company announced the beta version of its Meta AI assistant, says that "a combination of sources are used for training, including information that's publicly available online, licensed data and information from Meta's products and services."

The post says the models weren't trained on people's private posts or their private messages, though it adds "we may use the data from your use of AI stickers, such as your searches for a sticker to use in a chat, to improve our AI sticker models."

Likewise, Meta says it may use your interactions with AI features, such as "messages to AI chats, questions you ask and images you ask Meta AI to imagine for you." It may also save details and context about you from conversations you have with AI characters.

Meta execs have talked about the company's approach to AI training before.

Meta's chief product officer, Chris Cox, said earlier this year that Meta's text-to-image model, Emu, can produce "really amazing quality images" because Instagram has many photos of "art, fashion, culture and also just images of people and us."

"We don't train on private stuff, we don't train on stuff that people share with their friends, we do train on things that are public," he said at Bloomberg's Tech Summit.

Besides user-generated content, Meta is also considering deals with news publishers for access to more training data, including news, photo, and video content, BI previously reported. AI competitors like Google and OpenAI have already made deals with many news publishers for training data.



Popular Right Now



Advertisement