Been used for decades: gaussian blur or pixelization to obscure the identity of an on-camera subject, usually a witness, a whistle-blower, someone who is taking a big risk to be on camera.
You can't count t on it anymore.
A recent paper came out demonstrating an AI using algorithms that can reconstruct a very, very blurred face with frightening accuracy and detail. No longer can you count on just a masked blur to keep your on-camera talent safe from ID. If someone with resources wants to - say, a government or large corporation... they can undo your blur and identify the speaker. So can some hacker with a lot of time on their hands.
What to do next? I don't know if this will help, but If I have to do this for a serious identity-preservation in future, I'm putting a motion-tracked, animated fake face over the real person's face, before I add the blur. On a maybe more practical level, rubber masks might become a thing.