Anthropic Reportedly Preparing to Watermark AI Text for EU Rules

In one line: Anthropic is reportedly preparing to add identifying watermarks to text generated by its AI in order to meet new EU regulations.

Key points

• Anthropic is said to be planning to apply watermarks to text produced by its AI models (Claude).
• The move is driven by the EU's AI regulation (the AI Act), which is tightening transparency and labeling requirements for AI-generated content.
• Watermarking text is technically harder than watermarking images or video, so its effectiveness and susceptibility to removal remain open questions.

Why it matters

Labeling requirements for AI output are likely to spread, starting with the regulation-forward EU. If a major AI company adopts text watermarking, it could signal that content authentication and regulatory compliance are becoming an industry standard.

Read more

• Anthropic Will Watermark AI Text for New EU Rules — AI News (Google)