Automorphic

Automorphic is a secure, self-improving language model framework that can infuse knowledge with just 10 samples. It also provides a model firewall called Aegis, which can detect prompt injections, prompt and PII leakage, malicious language, and more.