Anthropic is embedding an undetectable watermark in all text generated by its latest Claude models. The measure took effect in the European Union on August 2, 2026, and will be extended worldwide. The watermark, applied at the model level, persists during copying and may survive some text edits. Several open source projects, including one with 4,600 stars on GitHub, are already offering tools to remove it. The controversy is fueled by Anthropic’s past, which included a covert tracker in Claude Code that was removed in March.
Source: Read the original article
