Editorial opinion startupxo
It's interesting that watermarking is a technique that makes subtle changes to text word selection, so it remains even after copying or transferring. However, since detection rates are low for short sentences, math answers, and translations, and detection drops significantly even with just some word substitutions, I'm curious about how effective this will be in practical use. Also, the fact that it's only applied by default within the EU might serve as a measure of its global adoption speed. It will be important to observe how API users experience changes when they turn it on and off themselves.
Writing language: Korean