GLOSSARY
AI Watermarking
Invisible markers embedded in AI-generated content (or visible labels on it) so synthetic media can be identified later — a core piece of AI provenance.
Watermarking takes two forms: visible labels (the “AI-generated” badges on social platforms, metadata flags like C2PA) and invisible signals — patterns woven into pixels, audio spectra or token choices that a detector can later find. Google embeds SynthID in its models' outputs; OpenAI experiments with text watermarking; voice platforms watermark synthetic speech to fight fraud.
The honest state of the art: invisible watermarks are best-effort, not guarantees — cropping, re-encoding and determined attacks degrade them. Their real value is raising the cost of impersonation at scale and giving platforms a triage tool. For creators, the practical takeaway is disclosure norms: label synthetic media where a viewer could reasonably be deceived.