Glossary · Term

Watermark

Also known as: watermarking, AI watermark

Watermark is an invisible mark that identifies content created by AI. Mandatory measures are being discussed as a response to deepfakes.

Watermarking is a technology that plants a signal that is invisible to the human eye but can be recognized by machines in AI-generated images, videos, and text, allowing later confirmation that it is an AI product. Like the anti-counterfeiting pattern hidden in banknotes, it is like engraving proof of origin into the content itself.

As deepfakes and fake news have become social problems, it has evolved from the need for a means of distinguishing AI creations. Major AI companies are introducing technology that adds identification signals to image products, and discussions are underway in many countries to mandate the display of AI products by law.

However, watermarking is not an all-purpose solution. Signals can be weakened or intentionally removed through editing or recompression, and content created with an open model without a watermark has no mark in the first place, so it is more accurate to view it as one of several countermeasures.

✅ Why it matters

⚠️ Limits and debates

← View all glossary entries