Saturday, 18 July, 2026
The Cropping Loophole: Meta’s AI Detector Fails to Spot Trimmed Deepfakes
By TechShots Studio
Meta acknowledged that its preview AI image detection tool can fail to flag AI-generated content when photos are heavily cropped. A Reuters analysis revealed the system missed 55% of images created by Meta's Muse Image model once they were trimmed. The tech giant explained that while its invisible "Content Seal" watermark survives standard edits, significant cropping destroys the embedded signal.
Read full story at REUTERS