Who can request access to the watermark verification API?

Regulators, law enforcement, media, fact-checkers, independent researchers, educational organizations, EU civil society groups, and enterprises that need to verify watermarking for their own compliance can request access.

What technology does the watermark system build on?

The system builds on Google's SynthID text method, tweaking word-selection randomness to create a statistically detectable pattern that may persist through some editing.

What are the critics' concerns about the watermark?

Critics argue that if Claude picks synonyms based on a watermark key rather than meaning, text quality suffers. Trade publication "Artificial Lawyer" also flags transparency risks, since detectable AI fingerprints could become a problem where contracts ban AI use or during fee negotiations.

Large Language ModelsAI Regulation & PolicyAI Safety & AlignmentTHE DECODER1h agoSep 2, 2026, 06:00 JST2 min read

Anthropic opens Claude AI text detection to regulators, media, fact-checkers, and others

Send on LINE

Key takeaway

Anthropic is launching its watermark verification API. Approved organizations can check text for Claude's digital watermark.
The EU AI Act requires new Claude models to embed invisible watermarks.
Critics worry about quality and transparency risks.

3 Key Points

What happened
Anthropic is launching its watermark verification API, letting approved organizations check whether text contains a digital watermark from Claude.
Why it matters
The EU AI Act since August 2, 2025, requires new Claude models to embed invisible watermarks in their text output, making this detection system a compliance tool for regulators, law enforcement, media, fact-checkers, independent researchers, educational organizations, EU civil society groups, and enterprises.
What to watch
Anthropic plans to expand access over time. The system builds on Google's SynthID text method, tweaking word-selection randomness to create a statistically detectable pattern that, according to Anthropic, "may persist through some editing."

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

This launch comes as a direct response to regulatory requirements. The EU AI Act since August 2, 2025, mandates invisible watermarks in new Claude models' text output, pushing Anthropic to provide a verification mechanism. The system adapts Google's SynthID text method to create a statistically detectable pattern while claiming the watermark contains no user data and affects neither quality nor content. Critics, however, dispute the quality claim, suggesting synonym selection based on a watermark key could degrade text. "Artificial Lawyer" also highlights potential transparency risks where detectable AI fingerprints might complicate contracts banning AI use or fee negotiations, though it predicts the watermarks are mostly harmless. The API's targeted rollout to specific groups now, with plans to expand access over time, reflects a phased approach balancing compliance needs with broader deployment considerations.

FAQ

Who can request access to the watermark verification API?: Regulators, law enforcement, media, fact-checkers, independent researchers, educational organizations, EU civil society groups, and enterprises that need to verify watermarking for their own compliance can request access.
What technology does the watermark system build on?: The system builds on Google's SynthID text method, tweaking word-selection randomness to create a statistically detectable pattern that may persist through some editing.
What are the critics' concerns about the watermark?: Critics argue that if Claude picks synonyms based on a watermark key rather than meaning, text quality suffers. Trade publication "Artificial Lawyer" also flags transparency risks, since detectable AI fingerprints could become a problem where contracts ban AI use or during fee negotiations.

THE DECODERRead Original Article

Send on LINE