Instagram’s AI detection is a mess (again)

Instagram's "AI Content" labeling system is malfunctioning, automatically tagging non-AI-generated images while failing to label actual AI content. Users report false positives from minor edits using tools like Canva's Background Remover or iPhone Photos app, and images "poisoned" to disrupt AI training are also mislabeled. Meanwhile, tests show Meta AI's own generated images are labeled, but other AI tools are not consistently flagged. The author's experiments, including uploading both AI-edited and AI-generated images, found only Meta's own AI triggers labels. The inconsistency undermines the system's purpose of helping users trust content, leaving the detection methods unclear.

(Source：The Verge)