How Claude's text watermarking works
Future Claude models will generate text that contains a watermark. This is a way of determining the likelihood that Claude was involved in writing the text, and we, along with several other major AI providers, are implementing this change to comply with the EU AI Act. In this article, we share answers to some of…
Opening of the original on Anthropic News
Summary
Anthropic will embed undetectable text watermarks in future Claude models to indicate the likelihood of AI generation. This change complies with the EU AI Act. The method subtly alters word choices, using a secret key to modify the randomness that selects words. While undetectable to readers, the pattern allows verification. Anthropic states this process does not affect output quality, creativity, or readability, ensuring a seamless user experience.
Why it matters
This move by Anthropic addresses regulatory requirements like the EU AI Act, impacting all users of future Claude models. By implementing text watermarking, Anthropic joins other major AI providers in a push for transparency and accountability in AI-generated content. The chosen method, based on Google DeepMind's SynthID-Text, aims for imperceptibility to users while enabling post-hoc detection. Competitors are also exploring similar techniques. Future developments will focus on the robustness and widespread adoption of these watermarking standards.
Related: OpenAI: Public-Sharing Labels and Verbatim Field Egress in an MCP-to-A2A Agent Configuration: A Controlled Multi-Model Study · Google: AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy Probes · Microsoft: Responsible AI in 2026: How we are adapting for what’s ahead · Meta: Trump Administration Sides With OpenAI in New York Times Copyright Lawsuit · xAI: Ajeya Cotra – "This might be the clearest warning shot we ever get"
Rated middle: a real update, not a headline event.
- Black Box: The Chatbots | Spirals | Ep 1 – podcastThe Guardian AI · Web · Sep 3, 2026
- Black Box: The Chatbots | Spirals | Ep 1 – podcastThe Guardian AI · Web · Sep 3, 2026
- Black Box: The Chatbots | Spirals | Ep 1 – podcastThe Guardian AI · Web · Sep 3, 2026
- Four major AI models suffer rare overlapping downtimeArs Technica AI · Web · Sep 3, 2026
- US government sides with OpenAI on issue of training LLMs on copyrighted materialTechCrunch AI · Web · Sep 2, 2026
- Public-Sharing Labels and Verbatim Field Egress in an MCP-to-A2A Agent Configuration: A Controlled Multi-Model StudyarXiv cs.AI · Web · Sep 1, 2026
- Public-Sharing Labels and Verbatim Field Egress in an MCP-to-A2A Agent Configuration: A Controlled Multi-Model StudyarXiv cs.AI
- AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy ProbesarXiv cs.AI
- AI News: OpenAI Made a Massive Move Against NVIDIAMatt Wolfe
- I Built a FREE App That Runs Your Entire BusinessMatt Wolfe
- Defense-as-Skill: Evolving Runtime Guard Skill for Skill-Augmented AgentsarXiv cs.AI
- Improving our alignment and security practices Anthropic News
Questions people ask
- How does Claude's text watermarking work?
- The watermark subtly alters the randomness used to select words. A secret key modifies word choices, creating a detectable pattern without changing the output's meaning or quality for the reader.
- Will watermarking affect Claude's output quality?
- No, Anthropic states watermarking does not impact the content, creativity, or readability of Claude's text. Readers will not notice a difference between watermarked and unwatermarked outputs.
- Why is Anthropic implementing text watermarking?
- Anthropic is implementing this to comply with the EU AI Act, which requires AI providers to indicate the likelihood that AI generated specific content.
More from Anthropic News 40 more
Page generated Sep 3, 2026. Summaries are Subvolts' own; the story belongs to Anthropic News.