Re:Show HN: RAG knowledge base poisoning lab, 100% local
by aminerj on Friday March 13, 2026 @08:04AM
(#47363306)
Good question, the attack doesn't require direct DB access. Many RAG pipelines auto-ingest from web crawlers, RSS feeds, third-party docs, or user uploads. If an attacker can get a crafted document into any of those pipelines, they control what context gets retrieved. The threat is the ingestion surface, not the database itself.
Re:Show HN: RAG knowledge base poisoning lab, 100% local
by gmerc on Friday March 13, 2026 @10:31AM
(#47365010)
Insider threat. Every fucking large business has disgruntled employees, like Meta right now after finding out about Zuck's plan to flatten all roles to ICs
Re:Show HN: RAG knowledge base poisoning lab, 100% local
by skrebbel on Friday March 13, 2026 @01:18PM
(#47367118)
How is that different from "People with access to your CMS can put terrible lies on your company website"? The vulnerability is still "people who have access to things have access to things" but written maximally sciencey to hide the fact that there's no vulnerability.
Like why does it even matter what kind of page to craft when some company's AI bot source database is wide open? I simply don't understand this kind of post, they do lots of effort to suggest that this is a super big scary vulnerability but actually the "vulnerability" is:
> Each [automated pipeline into your knowledge base] is a potential injection path.
In other words, the tldr of this article is
- if your knowledge base is compromised
- then your knowledge base is compromised!!!!