"It's a conspiracy against genuine Claude users." "I'm canceling my paid subscription."
Anthropic's announcement that it would introduce watermarking for AI-generated text drew sharp backlash from some users, with many threatening to cancel their subscriptions. In response, the company moved to calm the uproar, saying the measure is required under EU law and that other AI companies face the same obligation.
Concerns over Anthropic's watermarking plan flooded SNS platforms including X (formerly Twitter) and Reddit on Saturday.
Users expressed outrage, calling the move "a conspiracy against genuine Claude users," while others announced they would cancel their paid Claude subscriptions.
Some worried that text they had written themselves — and only lightly polished or proofread with Claude's help — could be classified entirely as AI-generated content, putting them at a disadvantage.
Others pushed back against the critics, arguing that "the only reason to oppose this is if you were trying to deceive people with AI."
As the backlash mounted, Anthropic posted a clarification on its website, saying watermarking would be introduced not only for Claude but for AI models from other companies as well.
"Starting Aug. 2, the EU has required AI providers serving its market to label AI-generated content," Anthropic said, adding that "other major developers will also be introducing their own watermarking systems."
Anthropic also offered a detailed explanation of its approach in an effort to address user concerns.
The system compares word-choice patterns across a piece of text against Claude's own patterns to determine whether the content was AI-generated. That means text a user wrote and only lightly edited with AI assistance would not be flagged as AI content.
Conversely, the company said, even text originally generated by AI would likely go undetected if the user had substantially rewritten it.
Anthropic added that it would also release a tool allowing users to check for themselves whether a piece of content is identified as AI-generated, in the interest of transparency.
hanimomo@heraldcorp.com
