From
To
Pangram Labs reports that AI watermarking, meant to establish content provenance, is flawed as bad actors can easily bypass it, necessitating detection methods instead.