Meta has started training its artificial intelligence models using private photos that users have not uploaded to Facebook or Instagram. Previously, Meta relied on billions of public images shared on its platforms.
According to a report by TechCrunch, Facebook users attempting to post Stories recently saw prompts asking them to opt into “cloud processing.” This feature allows Facebook to regularly select and upload media from users’ camera rolls to the cloud to create AI-driven content such as collages, recaps, and themed edits for occasions like birthdays or graduations.
By enabling cloud processing, users agree to Meta’s AI terms, granting permission for the AI to analyze unpublished photos, including facial features, dates taken, and the presence of other people or objects. Meta also retains the right to store and use this personal data.
Meta has confirmed it has trained generative AI models on public content posted on Facebook and Instagram since 2007, using only posts from adult users. However, the definitions of “public” and “adult user” remain unclear. Unlike Google, which explicitly excludes personal photos from AI training, Meta’s current terms do not clarify whether unpublished photos accessed through cloud processing are exempt.
Meta has not responded to requests for comment from TechCrunch or The Verge.
Users can disable camera roll cloud processing in their settings, which will also delete unpublished photos from Meta’s cloud after 30 days. However, the feature’s framing raises concerns about involuntary use of private data, bypassing users’ conscious decisions to share content publicly.
Additionally, Reddit users have reported that Meta has offered AI-powered restyling on previously uploaded photos without users’ explicit knowledge. One user noted that Facebook had automatically applied a Studio Ghibli-style effect to her wedding photos without consent.