Meta said it's been using your public Instagram photos and Facebook posts to train its AI

Mark Zuckerberg speaks at a conference, a couple poses for a selfie in Hong Kong.
Meta says it's been using Instagram photos and Facebook posts to train its new AI.Andrej Sokolow/picture alliance via Getty Images and Vernon Yuen/NurPhoto via Getty Images
  • Meta said on Thursday that it's been using public Facebook and Instagram photos to train its AI.

  • Its new virtual assistant uses both public text and photos, a spokesperson told Reuters.

  • However, the new model didn't touch private messages and posts, he said.

Meta has been using public Facebook and Instagram posts to train its new artificial intelligence assistant, a spokesperson said.

Posts that were fed to the new AI include both text and photos, Nick Clegg, Meta's president of global affairs, told Reuters in a Thursday interview.

However, the AI hasn't touched private posts, like those shared only with friends and family, Clegg said, per Reuters.


Meta has some measures to exclude private details from public datasets used by the AI, Clegg told Reuters. The outlet's interview did not mention any specific measures used.

Meta also isn't using private messages on its platforms to train the new model, Clegg told the outlet.

"We've tried to exclude datasets that have a heavy preponderance of personal information," Clegg said, per Reuters. He added that Meta hasn't used sites like LinkedIn to train its AI due to privacy concerns.

Meta opened its virtual assistant, Meta AI, to the public on Wednesday in a Beta release.

The new assistant is meant to help users create their own digital stickers based on text prompts, edit photos with text instructions, and chat with AI personalities like the rapper Snoop Dogg acting as a "Dungeons & Dragons" Dungeon Master.

Meta AI was built using its AI language model, Llama 2, and a text-to-image model called Emu.

Emu was trained using photos from Instagram and Facebook posts, while Llama 2 used other public data sets, a Meta spokesperson told Insider in a statement.

Clegg added that he expects some people to legally contest whether training AI with copyrighted content is fair use, per Reuters.

"We think it is, but I strongly suspect that's going to play out in litigation," Clegg told the outlet.

Authors, artists, and developers have been suing AI companies and firms like Meta over concerns that their work is being used without their consent to train a technology that could undermine their careers.

Facebook and Instagram users own the content that they post as long as it doesn't infringe on someone else's intellectual property rights, per Meta's policies.

October 3, 2023: This story was updated to reflect a comment from a Meta spokesperson.

Read the original article on Business Insider