dailyai.report

A new analysis by Gabor Koos argues that text watermarks cannot definitively prove a document was AI-generated. Attackers can bypass these markers through simple paraphrasing or mixing human-written text. This fragility renders watermarking an unreliable tool for legal or academic forensics. Practitioners should treat watermark detections as hints rather than absolute evidence.