Mitigating prompt injections on Generative AI systems

Developers Doro Hinrichs and Heather Logan explain how easy it can be to hack GPT through malicious prompts and demonstrate their web app used for experimenting with different types of prompt injection attacks and mitigations.