Tag: llm
All the articles with the tag "llm".
Prompt injection through operational data
If you point a language model at your logs, tickets or alerts, an attacker who can write to any of those is writing part of your prompt. What that enables, and how to contain it.