Anthropic will expand its hidden watermarking system to several older Claude models beginning November 3. The company notified users on Friday that the technology will be added to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6 and Claude Sonnet 4.6.
The initial rollout covered only Anthropic’s newest Claude versions. The expanded system uses the same technology already deployed in Opus 5.5, according to the company. It embeds an imperceptible pattern in the model’s word choices to help distinguish AI-generated text from human-written content.
Anthropic says the watermark will not affect the accuracy or readability of Claude’s responses. It also will not include identifying information about the user or organization interacting with the service.
The move follows pressure from European Union legislators for AI-generated text to carry a hidden watermark. The technology is intended to help identify AI-produced material, amid cases involving AI-generated citations, books written with AI assistance and the use of AI in scientific peer review.
Whether broader watermarking will discourage people from misrepresenting AI-generated work remains uncertain, but the expansion will make the system available across more Claude versions.
Sources: Mashable.