Vol. I · 2026
RAGrets

No RAGrets, yet another AI security blog

About

What this is

RAGrets is a blog about AI systems that were configured wrong. Retrieval pipelines, model servers, agent frameworks, and the glue between them. Each post takes one misconfiguration, shows what it did, and shows the config that should have been there.

The knowledge section is the reference shelf behind the posts: one entry per component or attack, explaining what it is, what it trusts, and where it usually breaks. Posts are dated and stay put. Knowledge entries get revised when the picture changes.

The name is a pun on RAG. It is written by James Ericsson.