Facebook wants unpublished images on smartphones

In order to obtain good training material for its artificial intelligence (AI), the meta company has many sources at its disposal. After dubious file-sharing platforms for copyrighted books, the tech company may now also want to access camera photos on its users' smartphone memory.

This is supported by a pop-up dialog box reported by various Facebook users. According to tech magazine TheVerge, this appeared automatically when they opened their camera in the Facebook app to take a picture for their own social media presence.

Facebook wants photo folders

In the window, Facebook is said to have asked users for their consent for "cloud processing". This means uploading selected images from the smartphone's photo folder –, i.e. primarily images that users have actually taken themselves – to a Facebook cloud. The purpose of this is that Facebook can automatically evaluate these images to offer users automatically created collages, retrospectives or AI redesigns at presumably suitable times.

Facebook would like to be able to choose which images go into the cloud for this purpose. What is particularly surprising is that these images are not those that have been uploaded by users to a meta platform in any way. Facebook explicitly wants to search the smartphone memory and take the images of users that it deems appropriate.

Consent for meta AI included

Another controversial aspect is that anyone who clicks on "Agree" in this pop-up window also agrees to the terms of use for Meta AI. This allows Meta's AI to evaluate things such as faces, the date the photo was taken, people or objects in the images. This also gives Meta the right to store and reuse personal data.

When asked by TheVerge, Meta denied that the images obtained through "cloud processing" are already being used for AI training. However, Meta did not answer further questions about whether the company plans to do this in the future. TheVerge sees this as a sign that something like this is likely.

Meta already uses a lot of data for AI

It would be another step by Meta to secure user data for training its AI models. Since the end of May, the company has been automatically using all data that users have set as "public" in their Instagram or Facebook profile for this purpose, provided that the data subjects have not objected in good time. Some data is always considered public by Meta, including your name and username, your profile picture and avatars, as well as various activities in public groups, on public Facebook pages and Instagram channels. Private messages and data of users under the age of 18 should remain untouched.

(nen)