LearnTrack
AI and LLM
An AI and LLM test looks at an application built on a language model: what the model can be made to say, which tools and data it can reach, and whether one user's input can act with another user's access. It is not a review of how the model was trained or of the vendor that hosts it.
2 pages
What am I looking at
- I have a chat box backed by a language model Recon LLM app recon: mapping an AI feature
- I have an assistant that can call tools or read documents Recon LLM app recon: mapping an AI feature
- I have an input that makes the assistant act for another user Access Not yet written
- I have a confirmed AI or agent finding to write up Report How to write up an AI or LLM finding
Kill chain
-
Recon
- LLM app recon: mapping an AI feature The first hour on an AI feature is for mapping what the model can reach, not for trying to make it misbehave. The risk sits in the tools, documents and accounts behind the chat box, and those can be listed through ordinary use, the documentation and the client's own configuration.
-
Access
Not yet written.
-
Escalation
Not yet written.
-
Persistence
Not yet written.
-
Impact
Not yet written.
-
Report
- How to write up an AI or LLM finding An AI finding is a security bug only when the model's output crosses a boundary: another user's data, an action the user could not take, an action nobody confirmed. The write up rates it by what that action could do, not by how clever the input was, and recommends controls that hold whatever the model says.