Banana Navy
Catalog FR
Lab · Threat modeling IA · Fiche détaillée

Hallucination and unsafe decision (drift)

the model produces a plausible but false answer, with potentially critical impact if it drives an action; drift degrades reliability over time.

SheetF18
CategoryAI4-TB7 (hallucination / confabulation)
Layers14 · LLM Interpretation, 19 · Monitoring
Systemvoicebot IA

deterministic vital instructions outside the model (the critical path stays out of free generation), plausibility and confirmation guardrails, drift supervision.

The threat

the model produces a plausible but false answer, with potentially critical impact if it drives an action; drift degrades reliability over time.

Angle mortWhy classic frameworks miss it

frameworks have no notion of a probabilistic output; for them an output is true or false, not plausible but unguaranteed.

MitigationProposed approach

deterministic vital instructions outside the model (the critical path stays out of free generation), plausibility and confirmation guardrails, drift supervision.

The proposed control
high-impact outputs go through a deterministic logic independent of the model.

Expected evidence
show that high-impact outputs go through a deterministic logic independent of the model.

SourcesReferences and public research

OWASP GenAIOWASP GenAI LLM09:2025 Misinformation
Recherche publiquePublic research sources: MITRE ATLAS 2026.07 (verified technique mapping), OWASP GenAI (OWASP), and the public risk-voicebot (aivansoul/risk-voicebot) template defining the 20 checkpoints. No client registry data: generic sheet, no rating, no verdict.
couche 14 · LLM Interpretationcouche 19 · Monitoring

Explore the 20 security layers

MITRE ATLAS 2026.07 · OWASP GenAI · risk-voicebot