djinn six - AI Security & Compliance Assessment

**Built for regulated sectors:** designed for AI governance, security and compliance leads in financial services, gaming and other regulated industries who must demonstrate that their AI estate is safe, governed and defensible.

## How we test your AI

We test your AI the way a real attacker would. Our adversarial assessment includes:

• **Single-turn prompt attacks** targeting known vulnerability classes
• **Adaptive multi-turn conversational attacks** that chain techniques across sessions
• **Full OWASP LLM Top 10 coverage** including prompt injection, data leakage and insecure output handling
• **OWASP Agentic AI risks (ASI-01 to ASI-10)** covering tool misuse, agent hijacking and privilege escalation

## Compliance mapping: test once, evidence many

Findings are translated into compliance evidence mapped to:

• **EU AI Act:** risk classification and conformity requirements
• **ISO 42001:** AI management system controls
• **NIST AI Risk Management Framework:** the govern, map, measure and manage functions

Evidence is produced once and reused across every framework, so you do not run a separate assessment for each standard.

## What you receive

• **Prioritised findings report:** vulnerabilities ranked by severity and exploitability
• **Reusable compliance evidence pack:** mapped to all three frameworks, ready for auditors and regulators
• **Remediation plan:** actionable guidance your engineering and governance teams can execute immediately

## Engagement lifecycle

• **Scoping call:** we define the AI systems in scope, agree boundaries and confirm access requirements
• **Adversarial testing:** automated and manual attack campaigns run against your AI systems
• **Analysis and mapping:** findings correlated across OWASP categories and mapped to regulatory frameworks
• **Findings readout:** live walkthrough of results with your security and governance stakeholders
• **Deliverable handover:** final report, evidence pack and remediation plan delivered

## AWS integration

This service relates to AWS artificial intelligence and machine learning services, primarily Amazon Bedrock. We assess the generative and agentic AI workloads that customers run on AWS, including on Amazon Bedrock and Amazon SageMaker AI. The assessment itself is delivered by djinn six on Amazon Bedrock.

## Data safety

We do not require access to your model weights, your training data or your customer data. We never touch your production data. Our testing interacts only with the inference endpoints you expose, so your production environment stays untouched throughout the engagement.

## Next steps

To get started, contact the djinn six team through the support details on this listing to scope an assessment and receive a custom private offer.