Users Angry as Claude Outputs Can Now be Caught by AI Tools

Anthropic has recently started incorporating indiscernible, machine-readable watermarks into Claude model outputs, sparking a contentious discussion surrounding AI transparency and the ownership of AI-assisted content. The development is in response to new transparency obligations under the EU AI Act's Article 50, which entered into force on August 2, 2026.

Companies are now required to embed machine-readable markings in generated or manipulated content to make it detectable using AI tools. Anthropic has also committed to the EU's voluntary Code of Practice for ensuring compliance. Understanding Claude's Watermark Anthropic states that Claude models launched on or after August 2 include watermarking capabilities by default, while older models are being retrofitted with the technology.

The watermark seamlessly integrates into generated text and remains undetected by users. It is not altered by copying, pasting, or making minor edits to the text. The watermark applies system-wide across supported Claude products, including Claude, Claude Code, Claude Cowork, and the API. Controversy Surrounding Claude's Watermark The introduction of Claude's watermark has ignited criticism on platforms like Reddit.

One user, newly registered under three weeks, argued that the new system could unfairly expose ordinary users who utilize Claude for tasks such as reorganizing writing, summarizing content, or alleviating writer's block. More technically adept users might argue that removing the watermark evidence through rewriting or paraphrasing could be possible.

Conversely, other Reddit users countered that submitting Claude-generated work as wholly human-created could be misleading. Critics contend that the watermarking system is unethical, asserting that users provide the initial ideas, instructions, context, and repeated refinements, while Claude functions merely as a tool. Supporters of the watermark system argue that it is merely intended to identify AI involvement, not to claim ownership of the final work.

The Debate on AI Transparency Some criticism has been directed at the apparent contradiction of AI companies watermarking generated content when they are trained using vast amounts of existing human-created material. Supporters of the policy argue that people should have a means to discern when content has been generated or processed by AI.

Anthropic acknowledges the limitations of their watermarking system, noting that the watermark might not always be accurate. The watermark may remain on content that Claude has only proofread, translated, or summarized, while heavy editing, paraphrasing, translation, or combining the text with other writing could render the watermark undetectable.

In such cases, a detected mark would indicate that Claude may have processed the content rather than definitively proving Claude's original authorship.