Meta has introduced a new model, Muse Image, from Meta Superintelligence Labs, which will be trained on Instagram user content. All public profiles are now automatically included in the process of using photos and Reels to create new images, unless the user manually disables this feature.
What happened
Meta has implemented the Muse Image model, which uses public photos and Reels from Instagram for training and content generation. To limit this process, users must manually go to the "Sharing and reuse" section in their settings and disable the relevant options. It is important to note that the opt-out mechanism is not retroactive: images that have already been created based on user data will not be deleted.
Context
Unlike an opt-in model, which requires explicit user consent, Meta is moving to an opt-out strategy. This allows the company to populate datasets for its multimodal models as quickly and extensively as possible, turning user content into an accessible resource for AI training.
Why it matters for the industry
For the industry, this step signifies a shift toward aggressive data collection methods and the consolidation of "data scraping by default" practices among major tech companies. This creates a powerful barrier for competitors but simultaneously increases pressure on privacy standards and may lead to a shortage of "clean" data in the long term.
Why it matters for users
Owners of public Instagram profiles may now find that strangers use their tags to generate new images featuring them via Meta AI. To protect their privacy, users need to manually check their profile settings.
What is not yet known / limitations
There are varying assessments of the consequences: while experts see risks to privacy and compliance, some developers note the positive aspect of simplifying access to large datasets for content creation.
Sources
Author
Look at AI, Editorial Team