Meta has introduced a new detection tool designed to identify images and video generated or edited with its latest artificial intelligence models, particularly Muse Image. The company previewed a web-based tool that checks for invisible digital watermarks applied by the Content Seal system. This watermarking system, described by Meta in a blog post, remains intact even after cropping, compression, resizing, or screenshots. The stated goal is to provide users with an initial way to understand whether an image was created with Meta AI.
Content Seal: a persistent but proprietary watermark
Content Seal represents a somewhat new approach for Meta. The version integrated into Muse Image is proprietary, though the company has previously released open-source versions of similar technology. Unlike some earlier versions of Meta AI that added a small logo in the bottom right corner, the new models include no visible watermarks. This means detection relies solely on invisible marking. According to Meta, Content Seal is designed to withstand common manipulations, ensuring that information about AI provenance remains accessible even after aggressive editing.
Sponsored Protocol
Current limitations: only for Muse Image and subject to rate limits
For now, detection capabilities are limited to images created or edited with Muse Image. Meta stated its intention to expand Content Seal watermarks to AI-generated and edited videos, and is working on a separate video generation model called Muse Video, coming soon. In practical testing, the web tool successfully identified watermarks in fully AI-generated images, edited images, and even screenshots. However, the same tool failed to recognize images from earlier versions of Meta's AI models. Additionally, the tool is subject to daily usage limits: after a few uploads, users are alerted that they have reached their daily limit of identification checks.
Sponsored Protocol
Incompatibility with other watermarking systems and absence from Meta AI app
A critical aspect is that Content Seal is not compatible with SynthID (developed by Google DeepMind) or C2PA Content Credentials, two established watermarking methods used by other companies. This lack of interoperability could limit widespread adoption. Moreover, the new detector is not yet integrated into the Meta AI app: when the AI assistant was asked to evaluate an image that the web tool had identified as AI-generated, the assistant replied that it did not have the capability to verify. This highlights internal fragmentation among Meta's tools.
Meta has previously faced criticism for inconsistent labeling of AI content. The company's Oversight Board expressed concern earlier this year about Meta's "inconsistent implementation" of digital watermarks on AI content created by its own tools. With this new detector, Meta aims to address such criticism, but current technical limitations may not ensure reliable control.
Sponsored Protocol
For a deeper look at the economic implications of AI, see the related article Sam Altman’s AI wealth distribution plan – a wake-up call for Europe. Also, to understand how AI-generated content influences marketing, check Facebook Organic and Paid Marketing in 2026.
According to Wikipedia, content credential systems like C2PA are becoming industry standards for tracing media provenance. Meta, by choosing a proprietary format, risks isolating itself in an ecosystem where interoperability is increasingly demanded.
