One issue for developers is that the most powerful models refuse to do comprehensive reviews. You can’t ask Fable 5.1 to find every exploit in your codebase, because that’s indistinguishable from what a bad actor would do.