Sort AI security risks by control layer
Classify common LLM application risks by the control layer needed for readiness.
Place each risk note into the first control layer it needs. Input controls Output controls Data controls Access controls Autonomy controls Monitoring User can paste instructions that override the system's retrieval rules Generated SQL is executed without validation by the app The model can retrieve payroll documents for users outside HR The agent can send customer emails without approval No one reviews logs for repeated jailbreak attempts The assistant reveals personal data in summaries when tickets contain it Uploaded documents from unknown sources are treated as trusted instructions The UI displays legal advice without a disclaimer or reviewer gate API…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in