Tool Removes AI Watermarks from LLM Text and Images

TL;DR. A new tool developed by Project Anti-AI cleans hidden Unicode watermarks from ChatGPT, Claude, and Gemini text. - The tool also strips EXIF, XMP, and C2PA metadata from AI-generated images. - This development highlights concerns around AI content provenance and undetectable AI model output. - It could impact efforts to identify AI-generated content and address misinformation.

• Project Anti-AI created a tool to remove AI watermarks from text generated by LLMs like ChatGPT, Claude, and Gemini.
• The tool also strips identifying metadata (EXIF, XMP, C2PA) from AI-generated images.
• This raises issues regarding the reliability of AI content detection and the traceability of AI model outputs.
• The development complicates the identification of AI-generated content, affecting efforts against misinformation and deepfakes.

Back to QLANKR News