Meta made its own AI detection system. It should have just used Google’s
TL;DR
Meta has introduced Content Seal, an invisible watermark that flags images generated by its new Muse models. The launch was buried as a footnote in the announcement of its image and video generation tools. Meta's Oversight Board had called on the company back in March to deploy its own tools against deceptive AI content. Critics see Content Seal as a less accessible and less reliable version of Google's SynthID, given that established standards like C2PA Content Credentials already exist.
Nauti's Take
That Meta is building watermarking into its image models at all is progress - platforms with billions of uploads need machine-readable provenance signals. The problem: a proprietary system alongside SynthID and C2PA fragments detection further, and a marker only Meta can verify does little for newsrooms and tool vendors.
Teams building content pipelines should stick with C2PA and treat Content Seal as a supplementary signal.
Video
Sources