Meta shows detector for AI images and video made with Muse Image

Meta has demonstrated a preview version of a web tool that identifies images and video created with the Muse Image generator. The app detects the invisible watermarks the model automatically embeds in every file it generates.

How the detector works

The principle is simple. A user uploads an image or video to the web app, which analyzes the file for the invisible digital marking that Muse Image embeds during generation. If a watermark is found, the content is identified as AI-generated.

Invisible watermarks are one of the key technical approaches to labeling AI content. Unlike visible logos or text labels, they do not change how a file looks, but they survive standard manipulations such as compression, resizing and posting through social networks. That is why the approach is attractive: users get "clean" content, while platforms and regulators keep the ability to verify its origin.

The system's reliability depends on how well the marking withstands outside interference. Specialized watermark-removal tools may reduce detection effectiveness. Meta has not yet disclosed details about the robustness of its solution.

• Muse Image embeds an invisible watermark in every file it generates
• The detector checks images and video for this marking
• The tool is presented as a web app in preview status
• Meta has not announced a public release date

Why does Meta need a detector for its own AI?

Building a verifier for its own generative product is not a paradox but a logical consequence of regulatory pressure. Meta runs the world's largest photo and video platforms, Facebook and Instagram, and sits at the center of the debate over traceability of synthetic media.

The European AI Act requires synthetic content to be labeled. Similar bills are advancing in the US. In this regulatory context, building a detector is not a goodwill gesture but a baseline requirement for a platform of this scale.

A number of AI companies already support the C2PA standard (Coalition for Content Provenance and Authenticity), which embeds provenance metadata directly in the file. With invisible watermarks, Meta is moving in the same direction: it makes verification part of the infrastructure rather than an optional add-on.

What is Muse Image

Muse Image is Meta's image and video generator. Its key feature is that automatic marking is built into the model from the start rather than added after the fact. Every file that comes out of Muse Image already contains a watermark, and Meta has now shown a tool that can read it.

Pairing a generator with a detector from the same vendor signals that content traceability has become a product priority, not a reaction to pressure.

What this means

If Meta takes the tool to a public release, it would be one of the first widely available AI-content verifiers from a platform of this scale. For newsrooms, fact-checkers and moderation services, it could potentially make spotting synthetic media easier. The key condition: the system works only as long as generators consistently embed the marking and users lack simple tools to remove it.

*Meta is designated an extremist organization and banned in Russia.