Anthropic has begun embedding invisible watermarks in text and images generated by Claude, and there is no opt-out. The policy was announced on August 11 and applies to every Claude product worldwide, part of the company’s commitments under the EU’s Code of Practice on Transparency of AI-Generated Content.
Claude-generated writing now carries an imperceptible, machine-readable watermark woven into the text itself. It travels with the text when copied and pasted, and may survive light editing. Files Claude generates — PNG, JPG and SVG — get digitally signed provenance metadata based on the C2PA open standard, which signals that the file passed through Claude and lets anyone check whether it has been tampered with.
The marking is applied at the model level, so it appears no matter which product produced the text: Claude, Claude Code, Claude Cowork, the API, or Claude accessed through AWS, Google Cloud or Microsoft Foundry. Models released on or after August 2 carry it from launch; support for older models is in progress.
What the watermark does and doesn’t prove
Anthropic is upfront about the limits. A detected mark is a signal, not proof: people use Claude to proofread, translate or summarise their own writing, and the output carries a Claude mark even when the ideas are human. The absence of a mark proves nothing either — heavy editing, paraphrasing, very short passages, and screenshots or format conversion can all strip it. Detection tools for users and third parties are still in development.
That nuance has not calmed the reaction. Users on X and Reddit called the policy “hugely problematic” and declared the end of “the era of unprovable AI writing”. The loudest complaints came from people who use Claude to polish their own work, and from coders who worry that a cryptographic signature will degrade generated code. Radio host Erick Erickson summarised the proofreading grievance:
“I had ditched Grammarly for Claude for proofreading because it does a better job. But now the stuff I’ve written will be watermarked that Claude did the work. This is ridiculous.”
Erick Erickson, radio host and blogger
The policy exists because of Europe. Nearly 200 companies have signed the EU Code of Practice, including Meta, Microsoft and OpenAI. Google has watermarked AI images since 2023 and has since expanded to text, audio and video; OpenAI has had text-watermarking technology for years and has chosen not to ship it, per the Wall Street Journal. The last time the Claude community organised this loudly was over Anthropic’s pricing changes.
What it means for Claude users in the UAE
The marking applies wherever Claude is offered, the UAE included. Anthropic has been courting regional businesses — its Claude for Small Business bundle targets UAE SMEs — and any marketing copy, code or content produced through Claude now carries the machine-readable trail. Proofreading and translation output is marked too, so human-written Arabic copy polished with Claude can read as AI-processed. Detection is a signal, not a verdict.
Can I remove the watermark from Claude-generated text?
There is no official removal tool. Heavy editing, paraphrasing or rewriting can weaken or erase the mark, but its absence does not prove the text is human-written — older models, very short passages and unsupported formats produce unmarked output too.
Does Claude watermarking apply in the UAE?
Yes. Anthropic says marking applies to supported models wherever Claude is offered, worldwide, and covers the API and cloud partners such as AWS and Google Cloud. UAE users get the same marks as everyone else.
Why is Anthropic watermarking Claude output?
Anthropic signed the EU AI Act’s Article 50 Code of Practice on Transparency of AI-Generated Content, which obliges providers to make AI-generated content identifiable to computer systems.
