Meta has been using user posts since 2007 to train AI

Meta✴, which owns Facebook✴ and Instagram✴, has confirmed that it uses public user posts published since 2007 to train AI models. This statement was made during a government investigation in Australia. However, billions of users outside the European Union (EU) and Brazil who keep their posts public do not have the opportunity to opt out of AI training.

Image source: VEPN / Pixabay

During the Australian government’s investigation into the use of AI, Melinda Claybaugh, global director of privacy at Meta✴, admitted that the company has been using all public text and photo material posted by adult users on Facebook✴ and Instagram✴ since 2007. The admission came after persistent questions from Green Party senator David Shoebridge. However, Meta✴ does not provide the ability to delete already collected data, even if the user changes the privacy settings.

Meta✴, in its privacy materials and blogs, mentions using public posts and comments to train generative AI models. However, the details of this process remain unclear. In June 2023, in response to a request from The New York Times about the timing and scope of data collection, Meta✴ did not provide a specific answer, only noting that changing privacy settings would prevent future collection.

Particularly troubling is the use of data from users who may have been minors in 2007. Claybaugh stated that Meta✴ does not use the data of users under 18 years of age, but could not give a clear answer to the question of how adult accounts created by them as children are processed. Senator Tony Sheldon asked about scanning public photos of children on adult accounts. Claybo confirmed that such data is also used.

Unlike users in the EU, who have the right to opt out of AI training thanks to local privacy laws, and users in Brazil, which recently banned Meta✴ from using personal data to train AI, most of the billions of Facebook✴ and Instagram✴ users do not have this option . Claybaugh was unable to clarify whether the opt-out option would be given to Australian users or anyone else in the future, citing the uncertainty of the current regulatory landscape.

The lack of an opt-out option has drawn criticism from human rights activists and politicians. Senator Shoebridge noted that the Australian government’s failure to enact adequate privacy laws allows companies like Meta✴ to continue to monetize and exploit even photos and videos of children on Facebook✴. This statement points to a global problem: legislation does not keep up with the pace of development of AI technologies and data collection methods.