Anthropic is facing user backlash and subscription cancellations after implementing an imperceptible watermark in Claude-generated text to identify AI-processed content and comply with the European Union's AI Act.

Key Points

• Anthropic began applying invisible watermarks to text from supported Claude models globally on August 2.
• The watermark persists even when Claude is used only for minor tasks like proofreading, translating, or summarizing original human work.
• Users cite concerns over professional authorship, potential contract penalties, and the inability to keep AI-assisted work private as reasons for canceling subscriptions.
• Anthropic plans to release a free API allowing third parties to detect these watermarks in text.
• Competitors like Google and OpenAI also utilize watermarking technologies, such as SynthID, to distinguish AI-generated content.

Why it Matters

The controversy highlights growing tension between AI developers prioritizing regulatory compliance and professional users who fear that embedded markers could compromise their intellectual property or client relationships. This shift toward mandatory transparency may accelerate a move toward provider-agnostic workflows as users seek to avoid vendor lock-in and potential professional scrutiny.