If an attacker wrote the next email your agent reads, what could it make the agent do?
Agents hold credentials, call tools and read content an attacker can write. Fourteen questions across seven control areas, mapped to the AI agent security control set anchored to the OWASP Top 10 for LLM Applications. You get the gaps, the evidence that would close them, and the full assessment to take next.
Ready to see where you stand?
You'll get a computed readiness indicator across Inventory & identity, Tool authorisation, Untrusted content, Approval & limits, Isolation & egress, Audit & monitoring, Testing & supply chain, the evidence that would answer each gap, and the full assessment to take next. Any critical gap is called out on its own, even when the overall score looks strong.
Assessment methodology: aligned to the My CISO Partner AI agent security control set, anchored to the OWASP Top 10 for LLM Applications and the NIST AI Risk Management Framework. It reports a readiness indicator based on your answers; it does not classify any AI system under the EU AI Act or any other law, and it makes no claim that an agent is safe.
See your full results
Enter your email to unlock your readiness indicator, domain scores, top gaps, the evidence that would close them and recommended next steps.
- Overall readiness indicator and tier
- 7 domain scores with the AI agent security control set (OWASP LLM Top 10 anchored) mapping
- Your highest-priority gaps, critical ones flagged
- The evidence that would answer each gap
- Recommended next steps and an email copy
Your AI Agent Security Quick Readiness Indicator
Executive summary
Domain scores
Weighted: Inventory & identity 20% · Tool authorisation 15% · Untrusted content 20% · Approval & limits 20% · Isolation & egress 10% · Audit & monitoring 10% · Testing & supply chain 5%. Expand a domain to see the contributing questions, your answers and the AI agent security control set (OWASP LLM Top 10 anchored) references they map to.
Your highest-priority gaps
Evidence that would answer each gap
What an assessor, a customer or an auditor would ask to see. Having it is the difference between a readiness indicator and a verified result.
Recommended next steps
Assessment methodology: aligned to the My CISO Partner AI agent security control set, anchored to the OWASP Top 10 for LLM Applications and the NIST AI Risk Management Framework. It reports a readiness indicator based on your answers; it does not classify any AI system under the EU AI Act or any other law, and it makes no claim that an agent is safe.
Please note: This assessment provides a readiness indicator based on the information provided. It does not constitute legal advice, a conformity assessment, or a determination of an AI system's legal classification, which should be confirmed with qualified counsel.
Recommended full assessment: AI Agent Security Readiness Assessment
Agents act. They hold credentials, call tools, read untrusted content and take actions that used to need a person. Identity, least-privilege tool authorisation, injection defence, action approval and audit for agents are new controls that most programmes have not defined. 30 minutes to scope. No obligation.