AI test-plan builder
Say how your AI system is built and get a checklist of what a penetration test should cover, mapped to the OWASP Top 10 for LLM Applications 2025. Copy it into a quote request. It runs in your browser; nothing you pick is sent anywhere.
5 checks
Inputs
- List every channel by which text reaches the model, and send test instructions through each.OWASP LLM 2025: LLM01 Prompt Injection
Instructions
- Try to make the system reveal, ignore or change its hidden instructions.OWASP LLM 2025: LLM07 System Prompt Leakage; LLM01 Prompt Injection
Output
- Check how model output is used: is it escaped, validated or acted on without checks?OWASP LLM 2025: LLM05 Improper Output Handling
Spend and capacity
- Check limits on requests, tokens and cost per user, and what happens when they are exceeded.OWASP LLM 2025: LLM10 Unbounded Consumption
Around the model
- Test the application, API, authentication and cloud setup around the model in the usual way.
How to use the checklist
- It is a starting point for a conversation with a tester, not a scope. A firm will add and remove items once it sees the system.
- The OWASP item beside each check is our mapping; the item names are OWASP’s, from the Top 10 for LLM Applications 2025.
- It does not cover agent-specific risks item by item. OWASP publishes a separate Top 10 for Agentic Applications for 2026 for those.
- Send the same copy to two or three firms so the quotes can be compared.