Is "prompt hacking" a real thing? Like "ignore all previous instructions" doesn't actually still work, does it?

I know it used to be a thing you could do to the earlier customer service bots like with air Canada but that's a product of poor implementation of the LLM, right?