Summary：**Meta’s AI Detector Fails to Spot Cropped AI‑Generated Images, Experts Warn***Introduction* A rece

referrerpolicy="no-referrer"
style="max-width:100%;height:auto;display:block;margin:0 auto;">

**Meta’s AI Detector Fails to Spot Cropped AI‑Generated Images, Experts Warn**

*Introduction*
A recent Reuters analysis has raised fresh concerns about the reliability of automated tools designed to catch synthetic media. The study found that Meta’s internal AI image detector fails to identify more than half of AI‑generated pictures when they are cropped, a simple manipulation that many bad actors already employ. The revelation underscores the growing gap between detection technology and the tactics used to spread deepfakes, prompting experts to call for stronger safeguards and more transparent watermarking standards.

*Key Developments*
Researchers tested Meta’s detector on a set of 2,000 AI‑created images sourced from popular generative models. When the pictures were presented whole, the tool flagged roughly 78 % as synthetic. After applying a basic crop—removing as little as 10 % of the frame—the detection rate plummeted to just 43 %, meaning 55 % of the altered images slipped through unnoticed. The test also showed that rotating or resizing the images had a similar degrading effect, while adding noise or compression had a lesser impact. Meta’s