What Has Changed

According to a report by TechCrunch, Google has introduced a new setting that allows users to turn off the “visible watermark” on its AI-generated images and videos. Until now, Google’s AI-generated content had a visible overlay indicating that it was generated, but in the future, users will be able to hide this overlay at their discretion. However, the article clearly states that this setting change only affects the visible watermark. (Source: techcrunch.com)

Difference Between Visible Watermark and Invisible Identification Information

TechCrunch’s summary states, “Turning off this setting won’t affect invisible benchmarks used to identify an AI-generated file.” This means that disabling the visible watermark will not affect the invisible mechanisms used to identify AI-generated files. In other words, Google is implementing a two-layer structure where users can remove the visible overlay, but the file itself will still contain embedded identification information. However, the provided source information does not specify the algorithm or metadata format used for this invisible identification method. (Source: Ibid. techcrunch.com)

Implementation Details and Setting Procedures

The provided source information does not clarify which product or app menu allows users to operate this visible watermark removal setting, nor does it provide specific operational procedures for APIs or consoles. Setting key names and CLI commands are also not mentioned, and there is no link to official documentation. Currently, the only primary information available to engineers is the TechCrunch article itself, so those who want to try the setting will need to start by checking this article. (Source: Ibid. techcrunch.com)

Summary

  • Considering the introduction of a design that leaves the on/off of visible watermarks to users, when designing the UI for products that handle AI-generated content, it is possible to consider giving users the option to switch between display and non-display.
  • Understanding that visible watermarks and invisible identification information are maintained independently, systems that verify content authenticity can be designed with verification logic that does not depend on the presence or absence of visible displays.
  • Since API and setting key names have not been released yet, instead of rushing to integrate them into in-house services, it is possible to plan to wait for the release of official documentation or follow-up reports.