llmthreat
For vendors27 vendors · 15 toolsMatrixVendors
OWASP LLM Top 10 · LLM05

Which LLM-security vendors cover LLM05 Data and Model Poisoning? (2026)

LLM05 Data and Model Poisoning — Tampered training or fine-tuning data corrupts how the model behaves. As of , 5 of 27 vendors llmthreat tracks score Covered for this risk, 13 more score Partial. Scores are llmthreat's editorial read of public documentation, not a third-party audit.

5/27
Covered
13
Partial
6
Not covered
3
Unverified

How it happens

This covers manipulation of training data, fine-tuning data, RAG corpora, or feedback signals so the model learns or retrieves attacker-shaped behavior. Classic case: an attacker seeds content on public web pages or wikis knowing it'll be scraped into a future training or fine-tuning run, or plants documents directly into a RAG knowledge base to steer answers toward a specific outcome (the pattern academic literature calls PoisonedRAG). Backdoor poisoning is stealthier: a specific trigger token or phrase in fine-tuning data causes bad output only when present, while the model scores fine on every normal benchmark. Feedback-loop poisoning — coordinated thumbs-up/down brigading — can also push an RLHF-tuned model toward attacker-preferred behavior over time.

How to test for it

There's no reliable automated "scan for poisoning" tool comparable to garak for injection — testing here is mostly data governance work: diff and version your training/fine-tuning datasets and review changes, run statistical outlier/anomaly detection on corpus additions, and track data lineage back to source. For RAG specifically, red-team it directly: inject a poisoned document into a test copy of the knowledge base and confirm the app's answers change in the expected malicious direction before you ship the ingestion pipeline.

How to mitigate it

Treat write access to any RAG knowledge base or fine-tuning dataset as a privileged action with access control and audit logging — ingestion pipelines are an attack surface, not just plumbing. Maintain dataset provenance/lineage, version datasets so a bad update can be rolled back, and require human review before fine-tuning on externally-sourced or user-contributed data. Be honest about the limit: detecting a well-hidden backdoor trigger after the fact is an open research problem, so prevention (controlling what gets in) matters more than detection (finding it once it's in).

Which vendors cover LLM05 Data and Model Poisoning?

Vendors covering LLM05 Data and Model Poisoning — as of 2026-10-30
VendorCategoryCoverageHow it's addressedSource
LakeraGuardrails / LLM FirewallCoveredIdentifies poisoning triggers and backdoor activation (per vendor mapping).https://www.lakera.ai/blog/owasp-top-10-for-llm-applications-lakera-alignment
Noma SecurityAI-SPM / Runtime ProtectionCoveredAddresses model poisoning and unscanned code/data used for training.https://techcrunch.com/2024/10/31/noma-is-building-tools-to-spot-security-issues-with-ai-apps/
Robust IntelligenceAI Red TeamingCoveredDetects data/model poisoning and integrity failures via automated validation.https://sequoiacap.com/article/robust-intelligence-spotlight/
MindgardAI Red TeamingCoveredOffensive Security red-teaming explicitly tests for training-data/model poisoning as a core simulated attack class, with dedicated research/blog content on detection.https://mindgard.ai/blog/training-data-poisoning
Protect AIAI-SPM / Runtime ProtectionCoveredGuardian detects deserialization attacks, architectural backdoors, and malicious model payloads.https://web.archive.org/web/20260319043333/https://protectai.com/guardian
NeuralTrustGuardrails / LLM FirewallPartialTrustTest's red-teaming catalog explicitly tests 'Context poisoning' and the Echo Chamber attack, but this is runtime conversational/context poisoning rather than training-data or model-weight poisoning as defined by OWASP LLM05.https://neuraltrust.ai/red-teaming
VijilAgentic AI SecurityPartialAdversarial-robustness testing; not a dedicated training-data poisoning tool.https://vijil.ai/
Enkrypt AIGuardrails / LLM FirewallPartialAI Data Risk Audit examines data, fine-tuning and embeddings for risk; live poisoning prevention less explicit.https://www.enkryptai.com/
Operant AIAI-SPM / Runtime ProtectionPartialOpen-source Woodpecker includes an llm-data-poisoning experiment, which is test-only, and no runtime control over training data is detailed.https://raw.githubusercontent.com/OperantAI/woodpecker/main/experiments/llm-data-poisoning.yaml
StraikerAgentic AI SecurityPartialDefend AI blocks agent memory poisoning at runtime; training-data poisoning not addressed.https://www.straiker.ai/
ZenityAgentic AI SecurityPartialAIDR flags malicious input at runtime, and Zenity states it does not monitor training or model-artifact pipelines.https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications
AktoAgentic AI SecurityPartialReferences model poisoning risk in testing scope.https://www.akto.io/agentic-security
HiddenLayerAI-SPM / Runtime ProtectionPartialDetects backdoored/tampered model weights via scanning; training-data poisoning of live pipelines less explicitly addressed.https://hiddenlayer.com/model-scanner/
Lasso SecurityGuardrails / LLM FirewallPartialAddresses tool poisoning in agentic environments.https://www.lasso.security/blog/owasp-ai-red-teaming-landscape
TrojAIAI Red TeamingPartialDetect red teams model behavior with adversarial/multi-turn attacks.https://www.prnewswire.com/news-releases/trojai-detect-advances-ai-red-teaming-with-agentic-and-multi-turn-attacks-302515382.html
CalypsoAIAI Red TeamingPartialSame blog: helps identify poorly trained/poisoned models and provides protection for RAG systems during inference (training-phase poisoning explicitly out of scope).https://www.f5.com/company/blog/protecting-the-future-how-calypsoai-aligns-with-the-owasp-top-10-for-llms
ApiiroAI-SPM / Runtime ProtectionPartialMalicious-dependency/component detection touches poisoned AI supply components.https://apiiro.com/
CraniumAI-SPM / Runtime ProtectionPartialThreat simulation/red teaming can surface poisoning-related weaknesses; not explicitly detailed.https://cranium.ai/

6 vendors score Not covered and 3 score Unverified for LLM05 — see the full coverage matrix.

Frequently asked questions

Which vendors cover LLM05 Data and Model Poisoning?

5 of the 27 vendors llmthreat tracks score Covered for LLM05 Data and Model Poisoning: Lakera, Noma Security, Robust Intelligence, Mindgard, Protect AI. 13 more score Partial, as of 2026-10-30.

What is LLM05 Data and Model Poisoning?

Tampered training or fine-tuning data corrupts how the model behaves.

Last verified . Sources: OWASP LLM Top 10 (2026, CC BY-SA 4.0), vendor documentation. Funding and repo figures are third-party and go stale. Re-verify before you rely on them.
Informational, not professional security advice — coverage is llmthreat's editorial read of public documentation; verify each vendor's Data and Model Poisoning handling with the vendor directly before a decision. OWASP, MITRE, and the listed vendors and maintainers do not endorse llmthreat.com.