Tech Companies Expand Watermarks to Trace AI Content

Generative AI can produce convincing text, images, audio and video at scale, making visual inspection and conventional detection increasingly unreliable. The technology industry is therefore shifting toward provenance at the point of creation: C2PA Content Credentials record how a file was captured or edited, while imperceptible watermarks can survive when metadata is stripped. The push matters for news verification, fraud prevention and compliance with the European Union’s AI Act, whose Article 50 transparency rules took effect on Aug. 2, 2026.

On Aug. 11, 2026, Anthropic said Claude models released from Aug. 2 would weave imperceptible watermarks into generated text and attach signed C2PA provenance metadata to supported files. Google said on May 19 that SynthID had marked more than 100 billion images and videos and 60,000 years of audio, while its verification tools had been used 50 million times. Canon, also in May, began rolling out its C2PA-compliant Authenticity Imaging System in EMEA for the EOS R1 and EOS R5 Mark II, following technical testing with Reuters.