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

Google says its SynthID watermark can still be detected after compression, resizing, cropping, and repeated sharing-like edits.

The article contrasts SynthID with C2PA metadata and notes Google is partnering with companies like OpenAI, Runway, and Nvidia to expand adoption.

A test found the watermark remained detectable even after simulated reposting, suggesting stronger resilience than many AI labels.

But the piece argues watermarking can identify AI-generated media without proving it is true or stopping misinformation from spreading.