Meta has shifted its approach to training artificial intelligence by looking beyond the billions of public images uploaded by users on its platforms, Facebook and Instagram. The company is now exploring the use of billions of images that have not yet been shared.
According to a report by TechCrunch on Friday, Facebook users attempting to utilize the Story feature have encountered notifications prompting them to opt into “cloud processing.” This option would allow Facebook to regularly access media from users’ camera rolls to create automated content such as collages, recaps, and AI-generated themes for occasions like birthdays and graduations.
By accepting this feature, users consent to Meta’s AI terms, which provide permission for the AI to analyze unpublished images, including facial features, the date the photos were taken, and the presence of other individuals or objects. Users also grant Meta the right to “retain and use” this personal information.
Meta has openly acknowledged that it has utilized data from all content published on Facebook and Instagram beginning in 2007 to develop generative AI models. While the company asserts that it only utilizes public posts from users aged 18 and older, it remains ambiguous regarding what qualifies as “public” content and the definition of “adult user” as of 2007.
In contrast to Google, which explicitly states that it does not train generative AI models with personal data from Google Photos, the current AI usage terms from Meta, effective June 23, 2024, do not clarify whether unpublished photos accessed through “cloud processing” are excluded from training data. Meta has not responded to TechCrunch for comments, and Technology News has also sought clarification.
Fortunately, users have the option to disable the camera roll cloud processing feature within their settings. Doing so will lead to the automatic removal of unpublished photos from the cloud after a period of 30 days. However, this workaround, presented as a feature, raises concerns over privacy and relies on users’ decisions to actively post photos for public viewing. Additionally, TechCrunch uncovered Reddit posts that indicate Meta has been experimenting with AI-generated restyling suggestions on previously uploaded images, often without user awareness. For instance, one user noted that Facebook had modified her wedding photos with a Studio Ghibli style without her consent.
