Knowledge Network · Oct 6, 2026
Podcast Outtake: AI is moving too fast
Third-party AI risk is moving faster than most enterprise vendor risk management programs can keep up with. As organizations embed generative AI in customer support, coding, analytics, and internal search, many of those ...
Retrieval augmented generation (RAG) is a perfect example of why LLM security looks different from classic application security. RAG helps an LLM answer questions using enterprise context, but it also introduces a new class of failure: the model may not reliably distinguish instructions from data. If malicious content lands in a knowledge base, ticket, wiki page, or document repository, it can become an indirect prompt injection that steers the model to reveal sensitive data or take unsafe actions. This risk expands with agentic AI, where the model is not only generating text but also calling tools, querying databases, and executing workflows. The blast radius grows further when agents are overprivileged, because a single prompt injection can pivot from a narrow task into broader systems like HR records, salary data, or production environments.
Check out the full podcast: https://www.buzzsprout.com/2120821/episodes/19718679
From Third Party Threat Hunting™ (TPTH™)
Put it to work
Vendor AI risk
The free BEACON™ class: 19 minutes on the six categories of vendor AI risk, with a real case to work through.
Take the free class →Privileged access
The free VAULTS™ training: 31 minutes on privileged access review, with a worked review.
Get the free training →