Facebook wants to look into your private photos. It reveals itself when creating Stories

Meta, the owner of Facebook, is introducing a controversial new feature – it wants to gain access to photos on your phone that you have never shared online. While Meta has so far only used publicly shared posts to train its AI models, it is now venturing into the unexplored territory of your private images.

According to TechCrunch, users are seeing a pop-up window when creating a new Stories post, asking for permission for “cloud processing.” At first glance, this innocent-looking feature promises to create “creative suggestions” such as collages, recaps, or AI photo edits. The problem lies in what is hidden behind this service.

What exactly is Facebook offering and what is it truly asking for?

Facebook explains in the pop-up window: “To create ideas for you, we will select media from your photo gallery and upload it to our cloud continuously, based on information such as time, location, or themes.” It further assures that only you will see the suggestions and your media will not be used for ad targeting.

However, by clicking “Allow,” you also agree to the Meta AI service terms, which allow the company to analyze your photos, including faces, using artificial intelligence. Meta gains the right to “store and use” any personal information you have shared with it to personalize its AI outputs.

Meta is vague about the use of your data

When The Verge confronted Meta with the question of whether these private photos would be used to train AI models, it received an evasive answer. Meta only stated that it is “currently” not training its AI models on these photos but refused to answer whether it plans to do so in the future.

This raises legitimate concerns, especially considering that Meta recently admitted to using all content published on Facebook and Instagram since 2007 to train its AI models. Company spokesperson Maria Cubeta told TechCrunch: “Media from the photo gallery may be used to improve these suggestions, but they are not being used to improve AI models in this test.” It is important to note the phrase “in this test” – what will follow remains a question.

How does this feature work and what does it reveal about you?

Meta spokespeople clarified that agreeing to this feature gives the company permission to retrieve unpublished content from your photo gallery for the last 30 days. However, the terms also state that some AI “suggestions” related to topics such as weddings or graduations may also be created based on older photos.

Analyzing your photos includes facial recognition, identification of objects and people in photos, tracking the date and location where photos were taken, and other metadata. Based on this data, Meta then creates personalized suggestions.

Facebook wants to look into your private photos. It reveals itself when creating Stories Jakub Kárník News Jakub Kárník News

Some social media users have already reported cases where Facebook automatically created anime versions of their older photos using Meta AI, suggesting that the system actively processes content you have previously uploaded.

How to disable Meta AI access to your photo gallery?

If you have already accidentally enabled the feature (which is unlikely for now, as the new feature is being tested overseas) or want to disable it preventively, you can do so as follows:

• In the Facebook app, tap “Settings & Privacy” and select “Settings”
• Scroll down to “Photo Gallery Sharing Suggestions”
• Here you will find two options for sharing your photo gallery with Facebook, including the one called “Get creative ideas created for you by enabling cloud processing of your photo gallery”
• Turn off both options

Why should you be cautious?

Entrusting private photos to any AI service provider is risky, but in Meta’s case, concerns are warranted for two reasons:

1. History of privacy issues – Meta has a history of several scandals related to handling user data.

2. Unclear data usage terms – While Google, for example, explicitly states in its terms that it does not train its AI models on personal data obtained from Google Photos, Meta’s terms remain intentionally vague.

This feature is currently being tested only in the USA and Canada, but it is likely that if successful, it will be rolled out globally. This is another example of how technology companies are balancing on the edge of privacy in an attempt to acquire data for their AI systems. At a time when publicly available data on the internet is no longer sufficient, they are turning their attention to our private storage.

If you care about your privacy, we recommend being extremely cautious when granting similar permissions and regularly checking the privacy settings in your applications.

Do you pay attention to privacy in AI applications?

Source: TechCrunch, The Verge