Tertius News is an AI-native newsroom: an AI model reads the linked source articles below and extracts what each outlet reported, so you can compare their coverage side by side. How this works →

New information became available — this story may be updated shortly.

Google introduces an AI watermark toggle

Google is rolling out a toggle that lets users disable visible watermarks on AI-generated images, videos, and music from its Gemini ecosystem. The move balances creative flexibility with digital provenance, as invisible SynthID watermarking and C2PA metadata remain embedded. The feature is unavailable in regions where law requires visible watermarks.

By Tertius News AI Desk 2 outlets · 2 articles Version 1 Coverage Aug 17–23, 2026 Published 3d ago

Google lets users hide visible AI watermarks

Google has announced a change that allows users to disable the visible watermark on AI-generated content produced by its Gemini ecosystem. The toggle applies across images, videos, and music generated by the Nano Banana, Omni, and Lyria models, according to one report. Users can turn off the visible emblem via the media settings in Gemini and in Google's video editor Flow, as reported by TahawulTech.com.

The move is designed to strike a pragmatic balance between creative control for end-users and digital provenance safety, according to Josh Woodward, VP for Gemini, who stated on X that the feature addresses that balance. The visible watermark toggle will not be available in countries where keeping visible watermarks is required by law, Woodward noted, per the same report.

Android Police, reporting on the feature's rollout for Gemini Nano Banana image generation, described the toggle as a global setting that appears as a "Media Watermark" option in the Settings menu. Turning it off once prevents the watermark from appearing on all images generated using Gemini until it is turned back on, the outlet reported. The publication found the feature may be off by default, as it was for its author.

Deeper provenance checks remain active

While users can hide the visible watermark, Google emphasized that underlying security frameworks remain fully active, as reported by TahawulTech.com. The invisible SynthID watermarking and the Coalition for Content Provenance and Authenticity (C2PA) open-standard metadata will continue to embed into all media generations, ensuring content origin can still be verified across Search and third-party validation tools.

Android Police provided additional detail on these technologies. C2PA is an open industry standard by the Coalition for Content Provenance and Authenticity, with metadata that is cryptographically signed and separate from EXIF, embedded by camera manufacturers like Canon, Nikon, and Pixel. While tamperproof, the publication noted, C2PA data is fragile; removal can be as trivial as a watermark, often unintentional via editing apps. Adobe was a founding member of C2PA, and Photoshop only preserves C2PA if Content Credentials is enabled.

SynthID is a watermark embedded in pixels, announced by Google DeepMind in 2023, and is resilient, surviving crops, filters, compression, and degrading only when cropped to about 10%, according to Android Police. Detection uses a Bayesian probability scale rather than a simple yes/no. Google also launched a SynthID Detector portal, waitlisted for media professionals, and users can ask Gemini directly if an image was AI-generated, tagging @SynthId for best results, the outlet reported.

Why the visible watermark was considered expendable

Android Police's report, which was largely opinionated, described the visible Gemini watermark as "security theater," arguing that it was trivial to remove. A 30-second generative fill or repair in Photoshop was all it took, and a cottage industry of reverse alpha-blending tools sprang up explicitly to remove Gemini watermarks, the publication stated. Many users removed the watermark for aesthetic reasons, not to hide the image's provenance, it added.

The report noted that the watermark's removal was an acknowledgment of what was already true: the visible logos were an "unreliable narrator" compared to the more robust SynthID protection built into the pixels.

Google also open-sources metadata validation library

In addition to the watermark toggle, Google is open sourcing a developer library named Credentio to support localized metadata validation in external applications, as reported by TahawulTech.com. The decision follows Anthropic's recent introduction of mandatory invisible text watermarking across its Claude platform to align with regulatory requirements under Article 50 of the European Union AI Act, according to the same report. While Anthropic's approach embeds machine-readable markers directly into text strings and media files worldwide, Google's latest policy emphasizes making front-facing watermarks optional to better serve creative workflows without sacrificing underlying detection capabilities.

Rollout

Google plans a phased rollout of the watermark removal toggle across its consumer and developer surfaces in the coming days, with integration into Google Search expected shortly, as reported by TahawulTech.com. Android Police noted the feature is still rolling out, so not all users may have it yet.

How each outlet told it

Android Police as of Aug 23, 2026, 1:17 PM UTC

Framing: Headline is informal and opinionated: 'Google let me remove Gemini's watermark because the real one is baked into the pixels'—highlights the author's personal experience and downplays the visible watermark as superfluous. — Opinionated and enthusiastic; uses phrases like 'thrilled', 'security theater', 'pointless', 'unreliable narrator.'

Facts Included:

Google announced a change allowing users to skip visual watermarks on AI-generated images from Gemini Nano Banana models.
The feature is rolling out and may appear as a 'Media Watermark' option in Settings.
It's a global toggle—turning it off affects all images until turned back on.
It may be off by default; the author found it off.
The watermark was trivial to remove: a 30-second generative fill or repair in Photoshop, or reverse alpha-blending tools.
Some users removed watermark for aesthetic reasons, not to hide provenance.
C2PA is an open industry standard by the Coalition for Content Provenance and Authenticity.
C2PA metadata is cryptographically signed, separate from EXIF, and embedded by camera manufacturers like Canon, Nikon, Pixel.
C2PA data is tamperproof but fragile; removal is as trivial as a watermark, often unintentional via editing apps.
Adobe was a founding member of C2PA; Photoshop only preserves C2PA if Content Credentials enabled.
SynthID is a watermark embedded in pixels, announced by Google DeepMind in 2023.
SynthID is resilient—survives crop, VSCO filter, Instagram crops, compression; degrades only when cropped to ~10%.
SynthID detection uses Bayesian probability scale, not simple yes/no.
Google launched a SynthID Detector portal, waitlisted for media professionals.
You can ask Gemini directly if an image was AI-generated, tagging @SynthId for best results.
Chrome extensions like Is This Image Real? can help check images.

Read the source article →

TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC

Framing: Emphasizes the introduction of an AI watermark toggle, framing it as a feature offering creative flexibility. — Neutral/descriptive, with phrasing like 'pragmatic balance' and 'ensures content origin can still be verified.'

Facts Included:

Google announced it will allow users to disable visible watermarks on AI-generated content from its Gemini ecosystem.
The toggle will apply across images, videos, and music from Nano Banana, Omni, and Lyria models.
Josh Woodward, VP for Gemini, stated on X the feature balances creative control and digital provenance safety.
Users can turn off the visible emblem via media settings in Gemini and Google’s video editor Flow.
Underlying security frameworks remain fully active when the visible watermark is off.
The visible watermark toggle is not available in countries where keeping visible watermarks is required by law.
Invisible SynthID watermarking and C2PA open-standard metadata will continue to embed in all media.
Google is open sourcing a developer library named Credentio for localized metadata validation.
The decision follows Anthropic's mandatory invisible text watermarking to align with Article 50 of the EU AI Act.
Google plans a phased rollout of the toggle across consumer and developer surfaces, with Google Search integration expected shortly.
Source cited as Mobile World Live.

Read the source article →

What was claimed, and who backed it

Each row is one claim, attributed to the outlet whose wording states it most clearly. Confidence rates how directly the source text states the claim — explicit and unhedged rates high; hedged, pieced-together, or internally inconsistent statements rate lower. It does not measure whether the claim is true. Status counts the distinct outlets we found asserting it — so a single-source claim can still show high confidence, and a multi-source claim can show medium. Every one of those outlets is named beside the status, so you can check the count against the list. For claims extracted before we began storing that list, the row says so: it names the outlet the claim is quoted from and states that we have not recorded which outlets backed it. Outlets wrote at different times, so a figure that evolves — a casualty count, for example — can legitimately differ between rows; check the "as of" time next to each claim's source.

Claim	Confidence	Status
ClaimGoogle announced it will allow users to disable visible watermarks on AI-generated content from its Gemini ecosystem.	ConfidenceHigh	StatusMulti-source (2 outlets) via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC, Android Police as of Aug 23, 2026, 1:17 PM UTC
ClaimThe toggle will apply across images, videos, and music from Nano Banana, Omni, and Lyria models.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimJosh Woodward, VP for Gemini, stated on X that the feature balances creative control and digital provenance safety.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimUsers can turn off the visible emblem via media settings in Gemini and Google's video editor Flow.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimUnderlying security frameworks remain fully active when the visible watermark is off.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimThe visible watermark toggle is not available in countries where keeping visible watermarks is required by law.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimInvisible SynthID watermarking and C2PA open-standard metadata will continue to embed in all media.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimGoogle is open sourcing a developer library named Credentio for localized metadata validation.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimThe decision follows Anthropic's mandatory invisible text watermarking to align with Article 50 of the EU AI Act.	ConfidenceHigh	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC
ClaimGoogle plans a phased rollout of the toggle across consumer and developer surfaces, with Google Search integration expected shortly.	ConfidenceMedium	StatusSingle-source via TahawulTech.com as of Aug 17, 2026, 11:46 AM UTC

Transparency & Audit Trail

2 outlets · 2 articles consulted: Android Police, TahawulTech.com 10 claims extracted Version 1 Written 2026-08-23