Anthropic implements invisible watermarking for Claude AI text

Anthropic has implemented an invisible watermarking system for text generated by its Claude AI models. The technology works by introducing subtle, statistical biases in word choices that are imperceptible to human readers but can be identified by specialized detection tools. This method leverages the inherent randomness in large language model outputs to create a detectable pattern.

While Anthropic stated the rollout is intended to comply with the EU AI Act, the decision has sparked significant controversy. Some users and experts argue that the implementation goes beyond what is legally required by European regulations. Critics note that the watermark can persist even through translations, summaries, or minor edits, which may impact users who utilize the AI for simple proofreading or stylistic adjustments.

In response to the new measure, some subscribers have reportedly canceled their memberships, and developers have begun creating tools to remove the watermarks. The company has opted for a global rollout, stating that the system cannot be applied selectively by region.