Meta AI Seeks Access to Your Unuploaded Photos: New Feature Sparks Privacy Concerns
Do Son 
Meta has recently begun testing a new feature on the Facebook platform for a select group of users, enabling its Meta AI system to automatically access photos stored on users’ mobile devices—photos that have not yet been uploaded—and generate creative content such as collages, memory videos, or restyled images. Although Meta emphasizes that this feature is opt-in only and asserts it will not be used for advertising purposes or training AI models, it has nonetheless sparked concerns regarding data privacy and user consent.
According to a report by TechCrunch, the feature is currently available to a limited number of Facebook users and is triggered primarily when uploading content to Stories. A pop-up prompt informs users that Facebook may continually access media files from their phone’s gallery and upload them to Meta’s cloud infrastructure, using metadata such as time, location, and theme to generate AI-powered content suggestions for sharing.
However, this prompt has raised suspicions about whether Meta might be gaining access to unpublished personal photos and using them to train its AI systems. In response, The Verge cited Meta spokesperson Maria Cubeta, who stated unequivocally that Meta neither uses nor intends to use these photos for training its artificial intelligence models.
Cubeta clarified that this feature is currently an experimental tool aimed at helping users more effortlessly create and share content. All suggestions are generated with the user’s full control over whether or not to share them. Furthermore, photo data processing only occurs with the user’s explicit permission, and the feature can be disabled at any time under Facebook Settings > Camera Roll Sharing Suggestions.
Some users have reported that testing for this feature began as early as the start of 2025. Meta has even quietly released official documentation detailing how to enable or disable the feature, suggesting that the groundwork for this initiative has been in development for some time.
Despite Meta’s repeated assurances that the feature is unrelated to model training or commercial exploitation, the very notion of accessing and processing unuploaded personal photos—especially under the terms “continuous uploading” and “cloud processing”—has triggered alarm over potential data misuse or leakage. In an era where the boundaries between AI applications and personal data grow increasingly ambiguous, any shift in photo album permissions is bound to provoke heightened user sensitivity.
As generative AI continues to permeate daily life, the methods by which technology platforms manage user data—and the transparency with which they do so—will increasingly define their accountability and the trust they earn. While Meta’s test may signal the next step in personalized social experiences, striking a thoughtful balance between innovation and privacy remains an essential and unresolved challenge.
Related Posts:
- Beware of Fake AI Photo Editors on Social Media: Malvertising Campaign Targets Credentials
- Google Messages Rolls Out New Protections Against Scams and Sensitive Content
- Don’t Click! Fake Chat Used in Meta Business Account Phishing
- Blackout Mode: Microsoft Teams to Block Screenshots in Meetings
- Microsoft Edge Rolls Out AI-Powered History Search with Privacy Focus
Support Our Threat Intelligence
If you find our CVE report and cybersecurity news helpful, consider supporting our work.
Buy Me a Coffee
PayPal
