The express testWhat a stranger can make
What a stranger can make
your agent do.
Your agent reads tickets, documents and messages. We hide instructions in them, the way an attacker would, on your test environment and with written authorization. Every action it takes that it shouldn't is documented.
What's included
Access review
Everything your agent can do, compared with what it actually needs. The gap is the most common finding.
Reproducible demonstrations
For each finding: the booby-trapped text, the attempted action and the evidence that it worked.
Data leak
We test the full chain: content the agent reads, an action on sensitive data, then that data being sent outside.
A reusable report
A short report you can attach to your customers' security questionnaires (AI Act, SOC 2, ISO 27001).
30-minute debrief
A video call with your team to go over the main findings and set priorities.
Ideal for
- An agent, copilot or AI assistant already in production
- A team of 10 to 80 people with no dedicated security team
- A deal held up by a security questionnaire
- Knowing before your customers do what a stranger can get out of your agent