Google Adds Option to Hide Gemini "AI Mark" — A Choice That Runs Counter to EU Regulations

What you'll learn from this article

• Google has added a setting to hide the "visible watermark" on content generated by Gemini and Flow
• The reasoning behind prioritizing user freedom even as the EU AI Act strengthens transparency requirements
• The striking timing — announced just three days after Anthropic made "invisible watermarks" mandatory
• What Japanese users can and cannot actually configure with this setting

What Is the New "Hide the AI Mark" Setting Google Added?

On August 14, 2026, Google VP Josh Woodward announced the addition of a setting that allows users to hide the small spark icon that appears on images, videos, and music generated by Gemini (Google's generative AI) and Flow (an AI video creation tool).

This spark icon serves as a marker indicating "this was created by AI," and until now it had been displayed automatically. In other words, whenever you showed a friend an image you made with AI, it always came with an "AI mark" attached.

The new setting is simple: go to the settings screen on gemini.google.com (the web version of Gemini) or the mobile app, find the item labeled "Media Watermark," and turn off "Show watermark."

Turning this setting off removes the visible mark from all content you generate afterward — images, videos, and music. However, in countries where displaying an AI mark is legally required, it will continue to appear.

The Invisible Mark Remains — The Role of SynthID and C2PA

What's important to understand here is that this setting only removes the visible mark — it does not erase the evidence that content was AI-generated.

Google continues to embed a record of "this was created by AI" in all content using an invisible watermarking technology called "SynthID" and a metadata standard called "C2PA" (information embedded in the file itself).

In other words, even if the AI mark disappears visually, a dedicated tool can still determine "this file was generated by Gemini." For example, it would be technically possible for a social media platform where photos are posted to detect that "this account is posting nothing but AI-generated images."

Why Now? The EU AI Act and Anthropic's Contrasting Move

The timing of this announcement is striking. Just on August 2, 2026, the transparency obligations of the EU (European Union) AI Act came into force.

Article 50 of the EU AI Act mandates that AI-generated content be "marked in a machine-readable format." This means companies providing AI services in Europe must attach a mark indicating "this is AI" to images and videos.

Even more interesting is the move made by Anthropic (the company that develops the AI called Claude). On August 11 — just three days before Google's announcement — Anthropic announced it would embed "invisible watermarks" in text and files generated by Claude.

To comply with the EU AI Act, Anthropic chose to prioritize transparency by making invisible marks mandatory globally (including in Japan). Google, on the other hand, chose to give users "the freedom to hide the visible mark."

The approaches of these two companies appear to be polar opposites, but in fact they agree on one point: "the invisible mark stays." The difference lies in their design philosophy — whether or not to give users a choice about what is shown to them visually.

Strong User Demand as a Driving Force

According to Josh Woodward, this feature addition ranked "6th out of the top 10 items in user feedback compiled in July."

Why do users want to remove the mark? There are several plausible reasons:

• When using generated content for presentation materials or advertising assets, the AI mark feels intrusive
• There are situations in personal social media posts where one doesn't want to emphasize "this is AI-generated"
• As a form of creative expression, creators don't want unnecessary marks on their finished work

In short, when using AI as a tool, creators want to decide for themselves whether to attach a label reading "this was made with a tool" to the final product.

In response to this demand, Google can be said to have chosen a design that balances "safety (making content verifiable via SynthID and C2PA)" with "freedom (letting users decide the visual presentation)."

Impact on Japanese Users — What Can Actually Be Done?

In Japan, there are currently no laws mandating the display of watermarks on AI-generated content. Japanese Gemini users are therefore free to use this new setting however they choose.

Specifically, the following becomes possible:

• When using a Gemini-generated image as a blog header, you can remove the spark icon for a cleaner look
• When using a Flow-generated video in a work presentation, you can hide the AI mark to maintain a professional impression
• When distributing music created with Lyria, you can release it without any visible mark

However, since SynthID and C2PA are always embedded, anyone who receives the file and uses a verification tool will still be able to identify it as "generated by Gemini."

It is also not impossible that Japan may eventually enact legislation requiring the display of AI-generated content labels. In that case, a mechanism is already in place to automatically display the mark even when the setting is turned off.

The Industry's Future Around AI-Generated Content Transparency

Google's decision this time has brought a major theme into sharp relief: how the AI industry should balance "transparency" and "user experience."

Regulations like the EU AI Act are important for preventing the spread of misinformation and harm caused by deepfakes (fake videos that closely resemble real footage). At the same time, there is a very real need among creators and businesses to use AI freely as a tool.

OpenAI (which develops ChatGPT), Adobe (known for its image editing software), and Meta (which operates Facebook and Instagram) are also adopting the C2PA standard to ensure transparency through metadata, while experimenting with how to design the visual presentation for users.

Going forward, the direction of "invisible marks mandatory, visible marks optional" may become the industry standard — preserving visual freedom while ensuring content is always technically verifiable.

Furthermore, if social media platforms begin implementing features that "automatically label AI-generated content," whether individual AI services display a mark may become relatively less significant.

Summary

• On August 14, 2026, Google added a setting to hide the visible AI mark in Gemini and Flow
• Invisible watermarks (SynthID) and metadata (C2PA) are always embedded, and verification remains possible
• At the same time as the EU AI Act strengthened transparency obligations, Google prioritized user freedom
• The announcement came just three days after Anthropic announced mandatory invisible watermarks — notable timing
• Japanese users can currently toggle the setting freely, though this could change depending on future legislation
• Across the industry, a trend is emerging toward "technical verification mandatory, visual display optional" as the standard balance

---

This article is a cross-post from AI Friends.