IMPLICATOR
Claude customers are canceling paid subscriptions after Anthropic began applying an invisible watermark to text produced by supported models worldwide. The statistical mark can remain when Claude proofreads or translates a person's original work, even though Anthropic says detection shows only that Claude processed the text. Paying subscribers on the Claude Max tier, which starts at $100 a month, are leaving because of the watermark, while Anthropic says it has seen no uptick in cancellations.
What Changed
- Anthropic began applying an invisible text watermark to supported Claude models worldwide on Aug. 2, 2026, when the EU transparency code took effect. Users cannot opt out on any plan or API tier, including enterprise accounts.
- Four paying users interviewed by Business Insider canceled Claude Max, which starts at $100 a month, over concerns the mark survives when Claude only proofreads, translates or lightly edits their own writing.
- Anthropic says it has not detected an increase in cancellations since announcing the watermark. The company had 300,000 business customers as of September 2025, a figure widely expected to have grown since then.
- Anthropic plans to offer a detection API but had released no public detector or implementation details as of Aug. 17.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Paid users walk away
Business Insider interviewed four users who said the watermark prompted them to leave. Arturo Villarroya, who works in public policy, canceled Claude Max on Wednesday, Aug. 12.
Villarroya used Claude for coding, research and corrections at work. He worried the mark could survive when Claude merely checked spelling or fixed citations in writing he had produced. He is switching his coding work to Cursor and Grok 4.6.
Vladislav Rajtmajer, a freelance developer in the Czech Republic, ended his Claude Max subscription on Tuesday, Aug. 11. He uses AI for code reviews and translations. Rajtmajer called the policy a "signal to look at Chinese AI providers and learn to work with other models."
Richard Echols, an AI consultant in Georgia, left Claude Max on Aug. 12 after using Claude to prepare business documentation. Echols said, "Even if you edit your own work that you wrote, they add the watermark anyway." Software engineer Santiago Alessandro Rivera Martínez also dropped his paid subscription during the week of Aug. 10, saying he wanted the AI portion of his work to be portable across providers.
A mark without an opt-out
The system nudges some low-stakes word choices according to a secret key, creating a statistical pattern across longer passages instead of inserting hidden characters. Users cannot opt out on any plan or API tier, including enterprise accounts. The company applies the rule at model level.
Anthropic began applying the policy globally on Aug. 2, 2026, when the EU transparency code took effect. The company said it lacks a durable way to limit watermarking by region. It plans to offer a detection API, but had released no public detector or implementation details as of Aug. 17.
Get Implicator.ai in your inbox
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
Anthropic sees no cancellation surge
Anthropic says it has not detected an increase in cancellations since announcing the watermark. Those four departures are measured against Anthropic's 300,000 business customers as of September 2025, a figure widely expected to have grown since then. People regularly announce Claude cancellations on X, and some recent posts cited a separate controversy involving CEO Dario Amodei's wife.
The public posts remain self-reported claims, Anthropic's cancellation statement is the company's own account, and no public independently verified count of watermark-related cancellations exists.
Aadit Sheth, cofounder of executive communications firm The Narrative Company, said on X that audiences should be able to tell whether the words they read reflect a person's own thinking and that watermarks are one way to ensure that. In its internal testing, Anthropic says it found no impact on content, level of creativity or readability.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
Writers prepare to leave
Michael Lopp, who writes Rands in Repose and is preparing a new edition of Managing Humans, has used Claude Max since the plan launched. He wrote the book himself but used Claude for editing, research and data work. The policy has him preparing to move his writing elsewhere.
"My writing is my work, and Anthropic's current strategy is aggressively writer-hostile," Lopp wrote.
Frequently Asked Questions
When did Claude start watermarking text?
Anthropic began applying the policy globally on Aug. 2, 2026, the day the EU transparency code took effect. The company said it lacks a durable way to limit watermarking by region, so the rule applies wherever Claude is offered.
Can Claude users turn the watermark off?
No. Users cannot opt out on any plan or API tier, including enterprise accounts, and the company applies the rule at model level.
Does the watermark appear on writing a person wrote themselves?
It can. The statistical mark can remain when Claude proofreads or translates a person's original work. Anthropic says a detection shows only that Claude processed the text, not that Claude wrote it.
How many people have canceled over the watermark?
No public independently verified count exists. Business Insider interviewed four users who canceled, and Anthropic says it has not detected an increase in cancellations since announcing the watermark.
Can anyone check a piece of text for a Claude watermark?
Not yet. Anthropic plans to offer a detection API, but as of Aug. 17 it had released no public detector and no implementation details.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Related stories
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
