Meta AI Wants Access To Your Entire Photo Library

Meta is once again in the spotlight over data privacy, this time for a new AI-powered feature on Facebook that asks for permission to access your device's entire photo library—including pictures you haven't shared publicly.

A standard alert that Facebook users must approve before uploading images. Meta introduced another feature that asks users to accept the company's AI data policies for a service that recommends content to post, enhances photos, and generates images. (photo Isa Farfan/Hyperallergic)

What is Meta's New AI Feature?

Some Facebook users are now encountering a pop-up when they go to post a Story, asking them to grant Meta AI access to their camera roll for “cloud processing.” According to the prompt, the feature is designed to make sharing content easier by using AI to suggest creative ideas. These include "restyling" photos, grouping images by themes like birthdays or graduations, and creating personalized collages such as “travel highlights.”

The notification explains that Meta will periodically “select media” from your camera roll based on time, location, and identifiable themes. While it assures users this data will not be used for ad targeting, the request for such broad access to private, unpublished photos has raised significant concerns.

The Privacy Paradox: Meta's Stance vs. Its Terms

In a statement, Meta has asserted that it is not using photos from users' camera rolls to train its AI models in this current test. A spokesperson clarified, “Camera roll media may be used to improve these suggestions, but are not used to improve AI models in this test.”

However, there appears to be a contradiction in the fine print. To use the feature, you must tap “Allow,” which also means you agree to Meta’s AI Terms of Service. These terms broadly state that Meta can use “personal information” to “improve AIs and related technology.” The pop-up itself says, “Media and facial features can be analyzed by Meta AI. To create ideas, we’ll use info like date and presence of people or objects.”

This leaves the door open for future use. The company did not provide an on-the-record answer when asked if it would use data from unpublished images to train its AI in the future.

Facebook settings allow users to opt in or out of “camera roll cloud processing.” (screenshot Valentina Di Liscia/Hyperallergic)

A History of Data Use and Regional Differences

This development doesn't happen in a vacuum. Meta has a well-documented history of using publicly available user data to train its AI. The company has admitted to using photos and text from public posts for AI training since 2007, a practice that has already fueled privacy debates.

Notably, your ability to control this depends on where you live. In the United States, Meta was not required to notify users that their public posts were being used for AI training, and there is no simple opt-out option, as reported by the New York Times. In contrast, stricter privacy laws in Europe grant users the right to opt out of this data scraping.

Broader Implications for Artists and Users

The new feature taps into a wider, ongoing controversy about how generative AI models are trained. Artists have long expressed concern over AI companies scraping publicly available images from the internet to train models that can then replicate their unique artistic styles. Many argue this practice could be detrimental to their livelihoods, and some have even pursued legal action against AI companies.

While Meta's new feature is opt-in, it represents a new frontier in data collection, moving from the public square of shared posts to the private domain of a user's camera roll. For users, the key question remains: is the convenience of AI-generated collages worth the price of granting a tech giant access to your entire personal photo history?