Instagram's AI labels misfire again, flagging real photos

This story follows a similar episode in 2024, when Instagram's "Made by AI" label wrongly flagged photos with minor generative edits. Meta then said it would adjust its approach to reflect "the amount of AI used in an image," but it has not revealed what specific indicators it scans for. Now the same kind of confusion is resurfacing, and Meta has not responded to requests for clarification.

The problem appears to stem from vague detection methods, as even simple assistive editing tools—like background removal, which uses machine learning but not text-to-picture generation—can trigger the label. Canva says some of its tools were wrongly categorized, but reports persist. Meanwhile, actual AI-generated content sometimes slips through, as one journalist found that only Meta's own AI outputs reliably received the tag in tests.

For ordinary users, the result is a trust deficit: if real photos get flagged and some fake ones don't, the labels lose their purpose. The photographer and brands like Halsey's cosmetics company have publicly clarified that no AI was used, yet the tags remain. Meta's opaque process is leaving users to guess what will be labeled, which undermines the feature's stated goal of helping people spot synthetic content at a glance.