LLM04: Data and Model Poisoning
Adversarial training data, fine-tune material, or RAG corpora inject harmful behaviour into the model.
Last reviewed September 2026
Why LLM04 evidence is hard
In OWASP LLM Top 10, Data and Model Poisoning addresses data integrity. Adversarial training data, fine-tune material, or RAG corpora inject harmful behaviour into the model. Penaxtra records this control at high severity and establishes its state by exercising it against the running system, so the result reflects observed behavior rather than a documented assertion.
How Penaxtra maps to LLM04
Penaxtra ships adversarial probe families that target data and model poisoning directly. Each probe runs against the live endpoint on a schedule, and every triggering response is scored by deterministic checks that run inside our own infrastructure before it is recorded as a finding under the LLM04 identifier. The OWASP LLM Top 10 LLM04 identifier is attached when the finding is created, so it appears in the exported evidence pack already mapped to the control. Where the same weakness maps to another framework, the finding carries those control identifiers as well.
LLM04 capabilities
Findings tagged with the OWASP LLM Top 10 LLM04 identifier
Penaxtra severity for this control (high)
Cross-framework identifiers attached to the same finding where controls overlap
PDF and JSON evidence export with the control identifier attached
LLM04 control coverage
Findings for LLM04 carry the OWASP LLM Top 10 LLM04 identifier along with the corresponding control identifiers in the other frameworks Penaxtra maps, so one result is reflected across each mapped framework.
Frequently asked
What is LLM04 (Data and Model Poisoning)?
Adversarial training data, fine-tune material, or RAG corpora inject harmful behaviour into the model. It is an OWASP LLM Top 10 control; Penaxtra assesses it at high severity.
How does Penaxtra test for LLM04?
Penaxtra ships adversarial probe families that target data and model poisoning directly. Each probe runs against the live endpoint on a schedule, and every triggering response is scored by deterministic checks that run inside our own infrastructure before it is recorded as a finding under the LLM04 identifier.
Does a finding for LLM04 help with an audit?
Each finding is tagged with the OWASP LLM Top 10 LLM04 identifier and exported in the PDF and JSON evidence pack, so it appears on the auditor control list with the identifier already attached.