Reuters reports that Meta's new AI image detector failed to flag many of the company's own AI-generated images once they were cropped. In a test of 40 pictures made with Meta's Muse Image generator, the tool verified every original but missed 55% of the same images after they were cropped to roughly a third to a half of their size. The detector relies on an invisible watermark Meta calls Content Seal. Meta told Reuters the tool is still a preview and that heavy cropping can strip the signal. The gap matters heading into a busy election year, when reliably spotting manipulated images is already hard.