Security review
AI agent security review
You have a chatbot or an AI agent in production, or about to be. We attack it the way a curious user or a motivated attacker would, and tell you exactly what breaks and how to fix it.
What we test
In scope for every package.
- Prompt injection
- Direct, and indirect through documents, emails, web pages and tickets.
- Jailbreaks
- Getting around the rules the agent is meant to follow.
- Prompt and data leakage
- Extracting the system prompt, internal data or other customers' data.
- Tool and action misuse
- Making the agent send, write, delete or pay when it shouldn't.
- RAG poisoning
- Planting content in the knowledge base that changes what the agent says or does.
- Brand-damaging output
- Answers you would not want screenshotted.
Out of scope: network and infrastructure penetration testing, code review and model training.
Packages
Fixed prices, ex. VAT.
Basic
One chatbot with one knowledge base
€4,900
Standard
A chatbot with RAG, integrations and login
€9,500
Advanced
Agents with tools, internal copilots, AI features in products
from €16,000
What you get
With every package.
- Executive summary
- Findings with severity, mapped to the OWASP Top 10 for LLM Applications
- Reproducible attack paths
- A prioritised fix plan
- Free re-test within 30 days
How it works
- 01
Permission and scope
Written authorisation and an agreed scope before anything is tested.
- 02
Test
Preferably in staging.
- 03
Report and walkthrough
We go through every finding with your team.
- 04
Re-test
Free, within 30 days, once you have fixed what we found.
We only test systems you own or are authorised to test, and only with written permission.
Have an agent in production?
Reviews from €4,900 ex. VAT.