AI company Anthropic added an imperceptible watermark to some of its Claude models to comply with EU AI Act requirements, aiming to identify AI-generated text. The measure has drawn criticism from users and professionals who fear it will appear in normal text processing tasks, leading some to cancel subscriptions and spurring a surge in demand for tools to remove AI markers. Several developers responded quickly, launching open-source watermark removal projects within days, with one gaining over 14,000 GitHub stars. While supporters of attribution labeling oppose the black-and-white approach, legal experts note such tools do not explicitly violate existing laws unless used to misrepresent content. As Anthropic plans to expand the feature, conflicts between watermarks and removal tools are expected to worsen.