I’m assuming this refers to things analogous to dependabot on GitHub where maybe it automatically updates a library version reference and runs the tests and creates a PR if everything seems good, or similarly for fixing style issues or other stuff that is pretty trivial and has good test coverage.

When you maintain an open source project on GitHub you will occasionally get some open source automated bot that submits a PR to do things like this without you even asking, and I’m sure there’s plenty more you can sign up for or implement yourself.

I wouldn’t really call it AI, but it is automated. I agree with the parent comment that a journalist trying to push an angle would probably lump it in as AI in order to make the number seem larger.

It's common at most mega-corps like google. For example, if a utility function in an internal library was deprecated and replaced with a different function that has the same functionality. A team might write a script which generates hundreds/thousands of PR's to make the migration to the new function.

You don't want a single PR that does that, because that would affect thousands of projects, and if something goes wrong with a single one, the whole PR needs to be rolled back.