Facebook Begins Using Private, Unpublished Photos to Train Its AI

Meta is shifting its AI training strategy by aiming to incorporate user images that haven’t been uploaded to Facebook or Instagram. Recently, users encountered prompts when posting to the Story feature, asking if they’d like to enable “cloud processing.” This option allows Facebook to select media from users’ camera rolls for regular uploads to generate creative content like collages and themes, while analyzing unpublished photos. Users consent to Meta’s AI terms, permitting the analysis of media, facial features, and metadata. Meta previously admitted to using public data from these platforms since 2007, but lacks clarity on what constitutes “public” data. Unlike other companies, Meta’s terms have raised concerns about using unpublished photos for AI training, as users can opt-out of the cloud processing feature. Additionally, there have been reports of unexpected AI alterations to previously uploaded photos, indicating possible privacy invasions.

Source link