Data poisoning
Manipulating training, fine-tuning or retrieval data to degrade a model or plant behavior that chosen inputs trigger (OWASP LLM05:2026). In 2025 research about 250 poisoned documents backdoored models from 600 million to 13 billion parameters alike.
- Also called
- training data poisoning · model poisoning
- Defined in
- OWASP Top 10 for LLM Applications 2026
- Identifiers
- LLM05:2026 Data and Model Poisoning
Covered in
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.
Defend agent memory and context against poisoning (OWASP ASI06)
OWASP ASI06 Memory & Context Poisoning: how an instruction or false fact saved once in agent memory, conversation summaries or a retrieval store steers later sessions, users and agents, and the write, scope, provenance, expiry and recovery controls that contain it.
Turn NIST AI RMF, ISO/IEC 42001 and EU AI Act duties into engineering evidence
What the NIST AI RMF and its Generative AI Profile, ISO/IEC 42001 and the EU AI Act, as amended in 2026, ask of teams that build AI features, turned into engineering practices and one set of evidence. Not legal advice.