Text AI watermarks are inherently easy to remove

An analysis argues that watermarks placed in AI‑generated text can be easily eliminated. The author contends that any watermark scheme for text is vulnerable to simple attacks. Removing the watermark does not require sophisticated techniques. The claim challenges the effectiveness of watermarking as a detection method. The discussion highlights the difficulty of reliably marking AI‑produced content. It suggests that reliance on watermarks may give a false sense of security. The article calls for alternative approaches to identify AI‑generated text. Readers are encouraged to consider the limitations of current watermark technologies.