Anthropic adds invisible watermarks to Claude outputs, sparking user backlash on Reddit
Anthropic introduced invisible watermarks to identify AI-generated text from its Claude chatbot to comply with the EU AI Act, prompting criticism and debate among users on social media.
Anthropic has begun embedding invisible code in the text produced by its Claude chatbot to mark it as AI-generated, a step taken to meet the EU AI Act’s Transparency requirements. The change has drawn mixed reactions on Reddit, where some users argue the watermark could expose routine uses such as students reorganising paragraphs or journalists summarising documents, while others contend it merely enables detection of AI-generated content and mitigates associated risks.
One commenter, posting under the name visionode, described the measure as a "draconian conspiracy" that would penalise ordinary users. Another user called the watermarks "unethical" and "disgusting," questioning whether the tool should claim credit for work they felt they had largely performed themselves. Respondents countered that the purpose of the watermark is not to assign credit but to allow detection of AI-generated outputs, citing potential harms.
Other participants offered more measured criticism, noting the irony of watermarking content generated by models trained on large datasets that include other people's work. The discussion reflects broader tensions between regulatory compliance and user expectations for privacy and flexibility when using AI writing assistants.