Anthropic is adding watermarking to outputs from its Claude AI system, and tech blogger John Gruber criticizes the move. He describes the watermarking as “patently offensive” and raises concerns that it may affect how Claude writes, including how the system might prioritize traceability over the quality of the text.
The reports frame the debate around the purpose and impact of AI provenance tools. While watermarking is often used to help identify or trace AI-generated content, critics argue it can change user experience or writing style. In Gruber’s view, the watermarking approach signals a trade-off between marking content for identification and preserving natural language quality.
Across the cited coverage, the central points are that the watermarking change is being introduced for Claude, and Gruber’s reaction is sharply negative, focusing on potential implications for writing quality and system priorities. The outlets differ mainly in how they characterize the criticism, but they align on the core event and the nature of the response.
