Anthropic's Claude Watermarking Draws Criticism and Subscription Cancellations

According to a report from the blog Futurism, the watermark system works by making subtle changes in the AI's word choices across generated text, imperceptible to humans but forming a pattern detectable with a tool. Anthropic said it implemented the system in response to the EU's AI Act passed in 2024, which requires AI companies to mark content generated or edited by their systems.

TechCrunch noted that other companies including Google, Meta, Microsoft, OpenAI, and Synthesia have committed to adhering to the EU's code. However, Forbes reported that the news about Claude watermarking sparked a firestorm on X, with users panicking that AI-assisted writing will now carry a kind of permanent "scarlet letter." Forbes also wondered whether media companies are normalizing AI while at the same time stigmatizing its output.

Meanwhile, Google announced on Friday that it will now let users remove the visible watermark from AI-generated images, videos, and music created with several of its models.

Business Insider pointed out a further nuance: Anthropic said the watermark indicates Claude processed text, not necessarily that it was the original author. The company said marks can remain after Claude proofreads, translates, summarizes, or otherwise edits content, meaning a watermark shows Claude processed text, not necessarily that it wrote it.

In the days since the announcement, dozens of people have posted on X since Monday that they canceled their Claude subscriptions, citing the watermark. Vladislav Rajtmajer, a freelance developer in the Czech Republic, told Business Insider he canceled his Claude Max subscription on Tuesday, saying the watermark was the main reason. He said he uses AI for code reviews and translations and worried that an AI label on code shipped to clients could raise questions about authorship or trigger contract penalties.

Richard Echols, an AI consultant in Georgia, also said he canceled his Claude Max subscription on Wednesday, with the watermark a major factor. He said he uses Claude to create documentation for businesses and worried the mark could appear even when he had written the underlying material himself and used the chatbot only for edits. "Even if you edit your own work that you wrote, they add the watermark anyway," Echols said.

Anthropic is not the only AI lab using watermarks. Google uses its SynthID technology to watermark AI-generated content, and OpenAI uses SynthID for supported images and audio. Elon Musk's social media platform X also adds a "Made with AI" tag on content it determines to be AI-generated or manipulated.

Long-time Slashdot reader kmleon, a tech consultant and RPG researcher, reported another issue. He said he had built software using his own custom-built AI tool, only asking Claude to do some testing. He discovered that Claude sometimes decided, since it was involved on the edges, to take credit for all his work, claiming some parts were either "100%" or "Co-Authored/Created" by Claude.Ai / Claude Code. He said he only became aware of it recently after setting up scanners to catch Claude incorrectly watermarking content. He found it has sometimes been mistakenly over-zealous in taking credit for creating or co-authoring docs, code, images, videos, etc., with about a dozen instances spanning the past year, most appearing in the last few weeks.

On the brighter side, the report suggests watermarks could help AI stop training on AI-generated text.