Claude Will Put Invisible Watermarks On AI Text And Images—And The Internet Isn’t Happy

Aug 12, 2026 12:56 AM - 2 hours ago 1

Topline

Backlash to a scheme by Anthropic to commencement including an invisible watermark connected matter and images generated by the Claude AI exemplary intensified Tuesday arsenic users railed against the thought of AI usage being detectable successful their work.

A telephone surface displaying Anthropic's Claude.

Gado via Getty Images

Key Facts

The new policy, which brings Anthropic into compliance pinch European Union regulations, was announced Tuesday and will put embedded watermarks into media created by Claude.

Claude-generated penning will transportation an invisible watermark that will "travel pinch the matter erstwhile it's copied and pasted elsewhere,” and different files the AI creates, including images, will now see digitally signed metadata.

Exactly what that looks for illustration successful believe remains to beryllium seen: Anthropic says it will people much accusation astir really to admit the watermarks, and warned they won’t beryllium foolproof—text watermarks whitethorn beryllium erased if the penning is heavy edited and a file's metadata could vanish if the format of the record changes, a screenshot is taken aliases done different means.

The caller policy, which users cannot opt retired of, will use to each Claude models and products worldwide, including Claude Code and Claude Cowork.

The scheme was deed pinch accelerated backlash connected X and Reddit, wherever users called it “bulls—” and “hugely problematic,” and declared an extremity to “the era of unprovable AI writing.”

The bulk of grievances came from group who claimed to usage Claude only for proofreading matter they’d written connected their own, and coders, who said they worry adding a cryptographic signature to a codification will “degrade the output.”

CRUCIAL QUOTE

“I had ditched Grammarly for Claude for proofreading because it does a amended job,” power show big and blogger Erick Erickson posted to X. “But now the worldly I’ve written will beryllium watermarked that Claude did the work. This is ridiculous.”

TANGENT

The caller argumentation announcement launched an online statement astir who should beryllium credited for AI-generated work: the instrumentality that made it, aliases the personification who told it to? One Reddit personification launched a vigorous chat by claiming quality users should get in installments for AI model-generated content, not the machine: “I gave the instructions, context, decisions, and countless refinements, Claude was the tool,” they said. Commenters who disagreed said the watermark is not Claude “claiming credit, it’s acknowledging the contented is AI created.” Others based on to not in installments AI is unethical: “Claude created the activity and ALL you did was springiness instructions. If your supervisor came to you and gave you instructions connected a task and erstwhile you completed it, he/she past proceeded to return 100% in installments for it, you mightiness beryllium understandably upset. Why? because each your supervisor did was springiness instructions and you applied your knowledge and expertise and did the coding,” 1 commenter wrote.

Key background

The caller argumentation brings Anthropic into compliance pinch an EU regulation called the "Code of Practice connected Transparency of AI-Generated Content." It provides rules for marking and detecting AI-generated and manipulated content, and for labeling of deepfakes and different AI-generated aliases altered text. As of the extremity of July, astir 200 companies had signed the codification and agreed to comply, including Anthropic, Meta, Microsoft and Open AI. Google has been watermarking its AI-generated images since 2023 and has since expanded to text, audio and video. OpenAI besides uses invisible watermarking for images.

SURPRISING FACT

OpenAI has agelong had the tech to watermark its matter output from models for illustration ChatGPT, but has chosen not to merchandise it. Internal statement complete whether aliases not to deploy it has been going connected for years, according to the Wall Street Journal, but arguments astir mendacious positives, circumvention and concerns it will thrust users to competitors has won out—for now.