Designing AI agents to resist prompt injection
How ChatGPT defends against prompt injection and social engineering by constraining risky actions and protecting sensitive data in agent workflows.
Soutenez OpenAI en consultant la ressource originale
Lire l'article original