The Risks of Prompt Injection

The vulnerability exists when you take a carefully prompted, crafted prompt and concatenate that with an untrusted input from a user. In software built on top of large AI language models it can be subverted by a user injecting malicious input. The robots will take it over. They will have injected all the prompts already. So what a random set of like ridiculous guard rails. That's not that you can find your way around it. You'll need another AI product to build the guard rails.