This is probably one of the most annoying features in the YouTube history, even more annoying than the COPPA update. If you weren't expecting that from the quality offered by most AI dubbing programs (like ElevenLabs), it's actually not the case. It simply overlays a strange TTS voice over the video audio (without the voices and sometimes with strange noises), which is a very lax approach. Even the TTS voice mispronounces some things (e.g. the TTS voice mispronounces firmware as "Finji W") or it misinterprets a part of the video. I always saw that feature on almost every video. Also this thing is vectorized, so you can ruin it more easily.
C2A!