← Scenario library

The Logs That Lied

An attacker plants prompt injection in the audit trail — so the LLM that hunts them erases the evidence

Technique first revealed 23 Feb 2023

RAG Knowledge Assistant
UntrustedExternal sourcesYour systemquerysearch🧑User💬Chat / AppInterface🎛️Orchestrator /Agent Loop🔍Retriever📚Knowledge Store/ Vector DB📥IngestionPipeline🌐UntrustedContent🧩Prompt Assembly🧠LLM🌐Attacker(during
InstructionsDataActionsControl / decisionFeedback / logscrosses a trust boundary
👆 Click a component or flow to inspect
SetupStep 1 / 7

The AI-assisted SOC

A security team gets far too many logs to read by hand, so they let an AI assistant summarise each day and highlight anything that looks like an attack. Analysts trust its summary to decide what to investigate.

AI RiskAtlas is an educational model of how GenAI & agentic systems work and fail. Architectures and payloads are illustrative and simplified for learning — not operational guidance. Real-world cases are summarised from public reporting.

Sources & further reading →·Built by Shi Yuan ↗