Instagram’s AI detection is a mess (again)

Instagram's AI labeling system is under fire again as users report widespread inaccuracies in how content is flagged. Meta’s 'AI Content' label is reportedly appearing on images that were merely edited with basic tools, such as Canva’s background remover or minor blemish-fixing features, rather than being generated by AI. The issue stems from Meta's scanning of IPTC and C2PA metadata to determine if generative AI was used. However, the system appears unable to distinguish between minor retouching and full synthetic generation. This follows a similar controversy in early 2024 when professional photographs were labeled 'Made with AI' due to minor Adobe retouches. Meta previously promised to refine its approach to better reflect the 'amount of AI used,' but the current failures suggest the detection system remains inconsistent. This has led to concerns that the labels are making it harder for users to trust any content on the platform, as real photos are mislabeled while some fully AI-generated images slip through without tags.