Watermarking (AI)

The embedding of imperceptible signals or patterns into AI-generated content, or into model weights, that can be detected later to verify the content's origin or identify which model produced it. An important tool for content authenticity and accountability, enabling platforms, regulators, and researchers to trace synthetic content back to its source and detect unauthorized use of proprietary models.
See also: content authenticity.