Anthropic’s Watermark Expansion Leaves Bypass Loophole in Two Older Claude Models
Original headline: Anthropic's watermark rollout leaves a loophole on two older Claude models
Anthropic expanded its text watermark to four additional Claude models to meet the EU AI Act deadline. However, on the older Claude Opus 4.6 and Sonnet 4.6 models, developers can bypass the watermark by tweaking a single sampling parameter. Anthropic’s response was limited to a compliance warning.
The bypass loophole threatens the effectiveness of AI transparency safeguards required by regulation.
Research angleHow does the ability to bypass AI watermarks affect user trust, examined through the Technology Acceptance Model (TAM)?
This story could become your next paper
HJITM works with researchers from study design through surveys, analysis and submission to international journals in generative AI and technology management. Bring us your research question.
This item is an AI-generated summary of reporting by The Decoder. Please check the original article for accuracy. Copyright in the original article belongs to The Decoder.