Particle.news

Google Lets Users Hide Gemini’s Visible AI Watermark

The company requires invisible SynthID marks and C2PA provenance so automated systems can still verify Google-generated media.

Overview

  • Google on Friday, August 14, 2026 added a Settings > Media Watermark toggle in the Gemini app and Flow that lets users remove the visible corner “spark” from images, video, and music.
  • Google said every Gemini output will continue to carry mandatory invisible SynthID watermarks plus C2PA Content Credentials metadata even when the visible mark is turned off.
  • Gemini can now detect SynthID in images, video (visual and audio tracks) and surface Content Credentials, including flagging video segments that contain Google AI material.
  • Google open-sourced Credentio for local provenance validation and replaced Gemini 3.6 Flash with Gemini 3.7 Flash as a backend upgrade while signaling plans to extend these provenance tools into Search and Chrome.
  • Observers warn visible marks are easy to crop away and some commercial tests and services claim they can weaken or bypass SynthID, which means platforms and businesses will need stronger tooling and policies rather than relying on a single technical signal.