A while back I built an LLM Firewall with AWS Lambda, a proxy that sits between users and the model to catch prompt injection. It worked, but it meant writing custom code for every app and having zero visibility into AI services I didn’t own.
That’s …
