How Anthropic plans to watermark Claude's AI-generated text
- Anthropic recently introduced invisible watermarks to text generated by its Claude AI, implementing the system globally to comply with the European Union's Artificial Intelligence Act, sparking user backlash.
- Based on Google's SynthID technology, the system creates an invisible pattern by influencing token choices during text generation. Anthropic stated the method helps distinguish AI-generated material from human writing without affecting content quality.
- Professionals, including policy worker Arturo Villarroya and consultant Richard Echols, canceled their $100-a-month Claude Max subscriptions on Wednesday and this week, citing concerns that invisible markers could falsely flag original work.
- Anthropic maintains the watermark has a "negligible" impact on speed and cost, clarifying that detection only indicates Claude processed the text. The company plans to release a free API allowing third parties to verify watermarks.
- Expansion to older Claude models is planned for coming months as users increasingly explore alternatives like Cursor and Grok. Despite company claims of no statistically significant cancellation spike, social media backlash regarding authorship concerns persists.
71 Articles
71 Articles
Anthropic explains watermarks in Claude's texts to comply with the EU's AI Act. Critics say this can make writing worse and confuse revision with authorship.
The editor of the successful chatbot Claude has just announced a new method to identify the texts written by his system. But the details are lacking and the reliability is not certain, says Touradj EbrahimiA huge battle is about to be revived. For months, text detection systems written by artificial intelligence are on the market. They are more or less reliable, sometimes show absurd results, but also regularly manage to give relatively correct …
Claude AI Faces Wave Of Cancellations Over New Invisible Watermark Feature
Anthropic is currently experiencing a noticeable wave of subscription cancellations following the global rollout of an invisible watermarking system embedded in text produced and processed by its Claude artificial intelligence models. The initiative, implemented earlier this month, has provoked substantial backlash, particularly among professional users who rely on the platform primarily for auxiliary tasks such as proofreading, drafting revisio…
Coverage Details
Bias Distribution
- 50% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
