Hacking generative AI: Limiting security risk in the age of AI