Meta is testing a new feature on Facebook that asks users to grant its AI access to their unpublished photo libraries, sparking fresh concerns over photo privacy, according to TechCrunch.
Some Facebook users have reported seeing pop-ups requesting permission for “cloud processing” of their camera rolls. The feature, which uses AI to “restyle” photos, group images by themes such as “birthdays” or “graduations,” and generate “personalized creative ideas” like collages, claims it does not use the data for ad targeting.
Meta’s spokesperson told Hyperallergic that the feature is opt-in, suggestions are only visible to the user unless shared, and camera roll media “may be used to improve these suggestions, but are not used to improve AI models in this test.”
However, Meta’s AI Terms of Service—which users must accept to use the feature—state that personal information can be used to “improve AIs and related technology.” This apparent contradiction has drawn criticism, especially given Meta’s long-standing practice of training its AI models on public posts since 2007.
While US users cannot opt out of having their public posts used for training, European users have more robust privacy protections and can opt out.
Artists and privacy advocates have expressed concern that AI systems trained on public and private images can replicate artistic styles and potentially harm creators’ livelihoods. Meta has not confirmed whether the feature will expand to Instagram or WhatsApp, nor whether data from unpublished images might eventually be used for AI training.