AI Security Assessment.

What is in scope

• prompt injection, direct and through documents, web pages and tool output
• data leakage: other users’ data, system prompts, connected records
• agents and MCP: tool permissions, memory, multi-step attack chains
• the web and API layer around the model, with authenticated roles

How the assessment runs

1. Scope. The application, the roles, the test environment and the attacker positions that matter are fixed in writing before testing starts.
2. Test. Hands-on AI red teaming and AI penetration testing within the agreed engagement window, typically 1 to 2 weeks.
3. Report. Within 5 business days after the window: every finding reproduced, with impact, evidence and remediation, plus a summary for management. Confirmed critical issues are reported the same day.
4. Retest. One retest of the reported findings within 30 days of the report.

What you receive

• Reproduced findings, each with its impact and evidence
• Remediation naming the location and the change, or saying where the fix is a design decision
• A summary for management written for a reader who will not open the technical section
• Retesting of the fixes for reported issues, within the agreed scope

Written to be handed straight to whoever asked for it: security questionnaires, SOC 2 or ISO 27001 evidence, ISO/IEC 42001 and NIST AI RMF work, procurement or investor review.

Continue monthly

The assessment is the usual starting point for Continuous AI Security. The monthly plan then continues from this report, without a new baseline.

Who it is for

AI SaaS teams, agencies that build chatbots for their clients, and firms in finance, insurance and law: anyone whose AI talks to customers or touches their data, with developers but no one whose job is to attack it.

Terms

A fixed price, agreed in a written quote before any testing starts.