Google's SynthID watermark is hard to break, but it doesn't solve AI disinformation

Google says SynthID’s watermark can still be detected after heavy edits, compression, resizing, and sharing.

The article compares SynthID with C2PA metadata and notes Google is expanding the technology through partners such as OpenAI, Runway, Nvidia, and others.

A stress test found the watermark remained detectable even after repeated image degradation.

Still, the piece argues durable watermarking alone cannot solve the larger problem of telling real from AI-generated content online or stopping misinformation.