Google Allows Removal of Visible Watermarks on AI-Generated Content

CoinWorld reports:

Google is beginning to adjust the way AI-generated content is marked. The company announced that users will now be able to remove visible watermarks from images, videos, and music, but this will not affect the invisible SynthID watermark or the metadata related to the C2PA standard.

Visible Watermarks Made Optional
This feature will first apply to the Nano Banana, Omni, and Lyria models. Josh Woodward, Vice President of Google Gemini, stated that the relevant toggle will appear in Gemini and the video editing tool Flow, with support for Search to follow soon.

According to Google, users can decide whether to display visible watermarks in the "Media Watermark" option in settings. This feature will be gradually rolled out over the next few days.

Two Layers of Identification Information Still Retained
Google claims that this adjustment aims to strike a balance between creative freedom and content identification. While visible watermarks help indicate the source of content, they may also affect professional production and creative use scenarios.

Even with the visible watermark turned off, the system will still retain two types of identification information: the invisible SynthID watermark and the metadata based on the C2PA standard.

• SynthID: An invisible marker embedded in the content
• C2PA: An industry standard that records source and editing information
• Search: Will support related identification capabilities in the future

Google Synchronizes Open Source Verification Library
Google stated that users can still determine whether an image was generated by AI through Gemini or Search. This means the company has not abandoned content traceability but has shifted the focus from visible markers to more suitable technical solutions for platform and tool identification.

In addition to product adjustments, Google has also open-sourced a new library called Credentio, allowing developers to embed local verification mechanisms in their applications. This move means that external applications can also access the capability to verify related marking information.

Additional Information: Before this update was released, Anthropic sparked discussions by adding watermarks to text and files generated by Claude, a practice seen as related to EU regulatory requirements. In contrast, Google emphasizes a route of "optional visible markers, with underlying identification retained."