Meta introduced Content Seal in July — an invisible watermarking system for AI-generated images that embeds a "hidden provenance signal" detectable even after images are cropped, compressed, resized, or screenshotted. The system arrived as a footnote in Meta's Muse image and video generation announcement, responding to the Oversight Board's March call for Meta to employ its own tools against deceptive generative AI content.
Critics note that Content Seal functionally mirrors Google's SynthID, yet Meta developed its own system significantly later. Meta already participates as a steering committee member of the Coalition for Content Provenance and Authenticity (C2PA), promoting the separate Content Credentials standard alongside Google. SynthID has also been adopted by OpenAI, suggesting Google is willing to license its technology to rival AI providers.
Content Seal currently has notable limitations: detection is only available through a dedicated web tool Meta is testing, with no built-in detection capabilities integrated into broader platforms. The author argues Meta should have adopted SynthID rather than launching a parallel system.