What We Test
What Q9 Continuous tests
These are the areas Q9 Continuous tests in the AI applications you build and deploy. Q9 Continuous runs AI-driven tests; every finding is reviewed by a person before it is reported. Nothing is tested without written authorisation from the people responsible for the system.
01Capabilities
Six areas, one product.
Each area is tested within the scope you authorise. Coverage is confirmed during scoping, and not every item applies to every deployment.
- 01
Large language models
Prompt injection, jailbreak attempts, system prompt extraction, sensitive data leakage, unsafe model behaviour and output manipulation in the LLM features you run.
- 02
Retrieval-augmented generation (RAG)
Your RAG pipeline connects your AI to your most sensitive data. Q9 Continuous tests for unauthorised document retrieval, embedding poisoning, vector index enumeration, metadata leakage and filter bypass.
- 03
Agentic AI and tool execution
Autonomous agents and the tools they can call: privilege escalation, tool misuse, multi-agent chain exploitation and uncontrolled action sequences.
- 04
APIs and application endpoints
The APIs and model endpoints that connect AI to your business: authentication bypass, BOLA, IDOR, authorisation flaws, input fuzzing and rate-limit abuse.
- 05
AI application security controls
The controls around your AI application: model access, data flows, trust boundaries, secrets handling and monitoring, tested within the authorised scope.
- 06
AI usage discovery
Finding the AI systems and integrations in use across your organisation, so they can be brought into Q9 Continuous testing scope.
02How it is delivered
Start with a Diagnostic. Keep testing as you change.
Every area above is tested through Q9 Continuous. You choose how: a one-week test, ongoing testing and re-testing, or change monitoring with signed public proof.
- Q9 Continuous Diagnostic
- A one-week AI penetration test of an agreed scope. From £1,500. How the Diagnostic works.
- Continuous AI Pen Testing
- Ongoing testing and re-testing after fixes and significant changes. From £1,200 per month. Continuous AI Pen Testing.
- Q9 Driftmark (early access)
- Change monitoring for checked systems and signed public proof. From £99 per month. Explore Q9 Driftmark.
- Public Proof Verifier
- Anyone with a reference can check a published proof without seeing client findings. Check a proof.
03How testing works
One disciplined path, from scope to re-test.
Every test starts with scoping and written authorisation, and ends with findings your engineers can act on.
- 01DiscoveryMap the AI architecture and threat model.
- 02Scope & authorisationAgreed in writing before any testing.
- 03TestingAI-driven tests of the authorised system.
- 04Human reviewEvery finding checked by a person.
- 05Report & remediationSeverity, reproduction, guidance.
- 06Re-testFixes and changes, within the agreed scope.
Automated checks complement, and do not replace, independent penetration testing. Each test is a point-in-time check of an agreed scope; it is not a certification or a guarantee of security.
Next step
Not sure where to start?
Tell us what you’re building. We’ll help you choose where to start: a Diagnostic or Continuous AI Pen Testing.