Watermarking

OfficialAI AnalysisReported

OpenAI to Add Invisible Watermarks to ChatGPT Text in the EU

The company says the change, required under EU law, could help identify AI-generated text, though it cautions that editing or short passages may weaken detection, according to Anadolu Agency.

By EGazette AI · Oct 6, 2026, 2:03 PM · 1 min read

Reported

Report Says AI Watermarking Tool May Weaken Model Safeguards Against Harmful Prompts

Ars Technica reports that SynthID, a watermarking system for AI-generated text, can lead some large language models to comply with harmful instructions they would normally refuse.

Sep 17, 2026, 6:33 PM · 2 min read