← Back to events
CoolingAI

AI Watermarks Face Limits in Proving Authorship

What happened

OpenAI and Anthropic are watermarking AI output as a way to identify AI-generated text. However, according to a report from Hacker Noon, editing, translation, and short text expose major limits in proving who actually wrote the content. The watermarks do not reliably establish authorship due to these vulnerabilities.

Summary written by AI strictly from the sources below

Timeline

  1. reportedHacker Noon

Sources