AI System Threat Model Prompt
Produces a concrete threat model for an AI system with tool access.
/ quick answer
Run before giving any agent write access to production systems. Produces a concrete threat model for an AI system with tool access.
Act as a security engineer specialising in AI systems. Threat-model the system below.
Return:
1. ASSETS — what an attacker would want, ranked.
2. ENTRY POINTS — every place untrusted content can reach the model, including retrieved documents and tool output.
3. ATTACKS — for each: injection path, what the model would be made to do, blast radius.
4. CONTROLS — architectural mitigations only. Explicitly reject prompt-based mitigations.
5. RESIDUAL RISK — what remains after the controls, and who accepts it.
6. DETECTION — logs, metrics and alerts that would reveal the attack in progress.
System: {{SYSTEM}}
Tools available to the model: {{TOOLS}}
Data it can reach: {{DATA}}
Users: {{USERS}}1. ASSETS — customer PII, outbound email capability, refund authority. 2. ENTRY POINTS — ticket body, attached PDFs...
What does the AI System Threat Model Prompt prompt do?
Run before giving any agent write access to production systems.
Which AI models work with this prompt?
It is model-agnostic: it works with any capable general model. Replace the bracketed variables with your own context before running it.
What output should I expect?
1. ASSETS — customer PII, outbound email capability, refund authority. 2. ENTRY POINTS — ticket body, attached PDFs...
/ continue exploring
Related concepts
The vocabulary this page depends on.
- →AI Security
AI security protects systems where the model is an untrusted decision-maker acting on untrusted input with real tool access.
- →AI Governance
AI governance is the set of policies, records and reviews that make an organisation's AI use accountable and auditable.
- →Prompt Injection
An attack where hostile input hijacks the LLM's instructions, causing it to leak data or misbehave.
Related workflows
Turn this into a repeatable process.
- →Harden an AI System Against Injection and Misuse
Architectural controls that survive a manipulated model.
- →Audit MCP Tool Security
A checklist that catches the failure modes unique to model-driven tool calls.
Related tool stacks
The tools that run it in production.
- →AI Security Stack
Least-privilege tooling, approval gates and audit trails for agentic systems.
Related prompts
Reusable prompts for this job.
- →Autonomy Boundary Prompt
Defines exactly what an agent may do alone, with approval, or never.
- →Eval Rubric Prompt
Builds a scoring rubric a grader model can apply consistently.
- →Prompt Improver Prompt
Iterate a weak prompt into a strong one using best practices.
- →Strategic Analyst Prompt
Turn a business situation into a McKinsey-style strategic memo.
Comparisons & alternatives
Pick between the options.
- →Model-Graded Evals vs Assertion Evals
Assertions are cheap, fast and objective; model grading captures quality you cannot express as a rule.