AI RESEARCH

Designing AI agents to resist prompt injection

OpenAI Blog

How ChatGPT defends against prompt injection and social engineering by cons