Anthropic has outlined how it intends to add invisible watermarks to text generated by Claude. According to an announcement on Friday, the company’s text marking system is described as “a version of the SynthID-Text approach,” an open-source watermarking technology from Google DeepMind that produces detectable patterns based on wording probabilities.
Alongside the text watermarking, Anthropic is introducing C2PA support for images processed by Claude. Both measures are being adopted to satisfy the company’s obligations under the European Union’s AI Act, which requires that synthetic audio, image, video, and text carry machine-readable marks.
Why it matters
The EU AI Act’s transparency requirements mandate that AI-generated content be identifiable. By adopting an existing open-source watermarking method and the C2PA standard, Anthropic is moving to align Claude with these regulatory obligations.
Who should care
Developers and organizations using Claude in the European Union, as well as those tracking AI transparency and content provenance requirements, will want to note how these marking systems are being implemented.