Anthropic announced that future Claude models will generate watermarked text using a variant of the SynthID-Text technique to comply with the EU AI Act, alongside attaching C2PA content credentials to generated images and files.

Key Details

• Future Claude models will incorporate text watermarking that alters the source of randomness during word selection without degrading quality, content, or speed.
• Anthropic signed the EU Code of Practice on Transparency of AI-Generated Content in July 2026 along with approximately 190 signatories.
• Generated images and supported files will include C2PA content credentials cryptographically signed in the file metadata.
• Anthropic plans to release a watermark detection API and will roll out watermarking to older Claude models launched before August 2, 2026.

In This Story