tech
Claude users are canceling their subscriptions, citing Anthropic’s new AI watermark
One developer said they feared Anthropic's watermark could lead clients to flag their work as AI-generated and raise questions about authorship.
TL;DR
- Anthropic is implementing an "imperceptible watermark" in text generated by certain Claude AI models.
- The watermark is designed to be persistent, even through copying, pasting, and some editing, to identify AI-generated content.
- Dozens of users have canceled Claude subscriptions, citing the watermark as a primary reason, fearing it could negatively impact client work and raise authorship concerns.
- Users like Arturo Villarroya, Vladislav Rajtmajer, and Richard Echols expressed worries about the watermark appearing on content they used Claude for light edits or translation on, or even when the underlying text was their own.
- Anthropic states the watermark indicates Claude processed the text, not necessarily that it authored it, and that it can remain after tasks like proofreading or summarizing.
- Other AI labs like Google and OpenAI also employ watermarking technologies for AI-generated content.
- Anthropic's watermarking is partly to comply with the European Union's AI Act and will be detectable via a future API.
- Some see AI watermarks as beneficial for transparency and preventing AI degradation from training on AI-generated data.