Meta AI Labeling Issues

Meta’s new AI-image labeling system is facing fresh questions after tests found its detector did not reliably recognize images generated by Muse Image, even though Meta said those outputs would include an invisible Content Seal watermark. The finding adds another complication to Meta’s rollout of Muse Image, the first creative model in its Muse Spark family from Meta Superintelligence Labs. Meta has already made unavailable an Instagram feature that let people generate AI images by @-mentioning public accounts, after complaints that the default-on setting for public Instagram users raised privacy and consent concerns, including from SAG-AFTRA and CAA.