Mitigating prompt injections by building a custom Python interpreter. Very cool research here. –> CaMeL offers a promising new direction for mitigating prompt injection attacks
Kevin Schaul
Visual journalist/hacker covering AI