Instagram Mislabels Original Photos as AI Content, Misses Actual AI Images Again
Updated
Updated · The Verge · Sep 4
Instagram Mislabels Original Photos as AI Content, Misses Actual AI Images Again
3 articles · Updated · The Verge · Sep 4
Summary
Users over recent weeks say Instagram has wrongly attached “AI Content” labels to ordinary photos while leaving some genuinely AI-made or AI-edited images untagged, undermining trust in the platform’s disclosure system.
Canva’s Background Remover emerged as one suspected trigger, with Canva telling one user some assistive AI tools were being tagged as generative, though reports say Instagram still flags some such images even after Canva said it fixed the issue.
Other false positives appear tied to minor edits, poisoned images, and some iPhone photo adjustments, while tests cited in the report found Meta reliably labeled only images made with Meta’s own AI tools and missed several files carrying C2PA or SynthID signals.
Meta previously faced a similar labeling problem in 2024 and had promised to better reflect how much AI was used in an image, but it did not clarify its current detection methods or respond to requests for comment.