Meta’s AI image detection tool sometimes fails to identify images generated by Meta’s own AI system after they are cropped, according to an analysis by Reuters. The report says that while Meta’s detector is designed to recognize AI-made images using signals such as watermarks or other embedded indicators, simple post-processing can interfere with those signals. Reuters’ review indicates that cropping—an edit that removes portions of an image and changes its framing—can make the detector less reliable, including when the cropped image is still produced by Meta’s AI. Other sources covering the same Reuters findings describe the issue as a weakness in AI watermarking or detection mechanisms under basic transformations. This means that attempts to evade detection do not necessarily require complex alterations; relatively straightforward edits may be enough to reduce the detector’s effectiveness. The report frames the problem as important for deepfake identification and authenticity checks, because cropping is a common action in everyday image editing and sharing. The sources do not indicate a specific remedy in the articles reviewed.