Commentary: Pangram and Watermarking Are Still Meh

ruthstarkman · x · 2026-08-26

Ruth Starkman published a commentary arguing that current AI detection technologies, such as Pangram and digital watermarking, remain ineffective. The article discusses the limitations of these methods in identifying synthetic text and images, as well as the challenges they face in real-world applications.

Original post →

More from Safety

Safety channel →