Meta’s AI wants access to your ‘Camera Roll’—Here’s why you should be cautious

Meta is rolling out a new AI-driven feature for Facebook users in the US and Canada, designed to transform forgotten snaps from your photos into share-worthy creations. The catch? Instead of combing through your uploaded albums, the AI delves into your camera roll.

This optional feature asks users to “allow cloud processing to get creative ideas made for you from your camera roll.” Once granted access, Meta’s AI combs through your photos and videos — even those tucked between random screenshots or documents — to uncover what it calls “hidden gems.” The system then suggests creative outputs such as collages, edits, or highlight reels that can be saved or directly shared on Facebook.

At first glance, the idea seems harmless — a touch of AI magic to help users rediscover forgotten memories. But, as often with Meta, the finer details raise important privacy questions.

What happens to your photos once uploaded?

After opting in, Meta confirms that the feature continuously uploads media from your camera roll to its cloud servers for AI analysis and content generation. The company maintains that these files “won’t be used for ad targeting.” Still, the real concern lies in whether such data could be used to train Meta’s broader AI systems.

This isn’t the first time Meta has experimented with similar functionality. A comparable test took place in June, during which the firm assured users that unpublished or private photos weren’t used to train AI models — although it stopped short of guaranteeing that this policy would never change.

In simpler terms, Meta’s servers will store and analyse your photos to generate ideas, but the data will not feed its AI training systems unless you actively engage with or post the AI’s creations.

How long will Meta keep your data?

According to Meta, the feature will “select media from your camera roll and upload it to our cloud on an ongoing basis.” During the earlier testing phase, the company acknowledged that some uploaded data could remain stored for more than 30 days. However, it hasn’t yet confirmed whether that duration still applies.

It’s worth noting that last year, Meta admitted its AI systems had already been trained on all public text and photos shared on Facebook and Instagram by adult users since 2007. So, while the company currently claims to exclude unpublished photos from training, history indicates such boundaries could shift in future.