Why it matters

Manipulated training, fine-tuning or retrieval data can plant backdoors or bias that pass ordinary tests and trigger only on inputs the attacker chooses.

How to prevent it

  • Track the lineage of training and retrieval data, and restrict who can change it.
  • Validate and filter data before ingesting it, and keep versioned snapshots so that you can roll back.
  • Evaluate models with targeted tests for backdoor behavior, not only overall accuracy.

Guides on this site

Main subject · Guide

Protect training and retrieval data from targeted poisoning

Track who can influence datasets, keep evaluation independent and investigate targeted behavior changes without relying on aggregate accuracy.

References

  1. LLM04:2025 Data and Model Poisoning (OWASP, 2025 edition)
  2. OWASP Top 10 for LLM Applications 2026 (OWASP GenAI Security Project resource page)

List data from the publisher. A guide is listed as the main subject when the weakness or risk is what the guide is about, and as covering it when a dedicated section or worked example addresses it.