AI engineering glossary · AI ethics law product
What is Watermarking AI outputs?
Watermarking AI outputs (also called AI-generated content watermark): Embedding a hidden or visible signal in AI-generated text, images, audio or video so that the content can later be identified as machine-made. Such marks help fight misinformation, but they can be weakened by editing and are not foolproof.
Watermarking AI outputs: an example
An image model adds an invisible pattern to every picture it creates, which a detector can later use to flag it as AI-generated.
Related terms
More AI ethics law product terms
- Accountability (AI)
- Activation rate
- AI audit
- AI governance
- AI incident
- AI Incident Database
- AI risk assessment
- Algorithmic bias
Where you learn it
Watermarking AI outputs is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.