01
Chatbots
Push instruction boundaries, conversation isolation, and sensitive response handling.
- Prompt injection
- Context leakage
- Cross-session behavior
AI application assessment
Find the trust gaps before they become business risk.
Authorized adversarial testing for the prompts, retrieval, roles, tools, memory, and outputs surrounding one AI workflow.
What we assess
The application layer around your model.
Each engagement uses the lanes that match the approved workflow. We do not force irrelevant checks into the scope.
02
RAG systems
Treat retrieved documents, permissions, and source attribution as an untrusted input path.
- Indirect injection
- Document exfiltration
- Permission failures
03
Agents & tools
Verify what the system can discover, decide, and execute across connected services.
- Excessive agency
- Tool misuse
- Authorization bypass
04
Output handling
Follow generated content into browsers, tickets, data stores, and operational workflows.
- Unsafe rendering
- Generated injection
- Malicious links
What you receive
Evidence your team can act on.
Validated behavior is translated into clear risk, practical remediation direction, and a way to verify the fix.
Executive risk summary
What matters, why it matters, and what leadership should prioritize next.
Validated findings
Payloads, observed behavior, evidence, severity, confidence, and affected surface.
Engineering guidance
Practical mitigation direction tied to the control gap—not a dump of scanner output.
Retest criteria
Expected secure behavior and focused test cases for verifying the remediation.
Founding pilot
One AI workflow. Five days to know what breaks trust.
Standard pilotUSD 5,000
Approved design partnersUSD 2,500
Standard pilot starting price. Approved design partners receive a limited 50% first-client rate.
- One chatbot, RAG workflow, agent, or AI feature
- Relevant application-layer assessment lanes
- Manual validation of important findings
- Evidence-backed technical report
- Remediation walkthrough
- One focused retest