AI Watermark Removers Rise While Verification Remains Elusive
After Anthropic announced imperceptible watermarks in Claude outputs, a surge of watermark-removal tools and web services has appeared, but none of the claims about defeating the watermark are verifiable yet; some tools only strip metadata or non-content markers, while removing the watermark embedded in wording would require a second model and heavy rewriting; independent testing has found gaps and many offerings lack audits or licenses, raising concerns about supply-chain risk and EU transparency requirements as no public detector exists yet.
