Prompt injection
We systematically test how your AI handles untrusted input — both direct user input and indirect content from documents, emails, web pages or database records that the model processes.
Service
You can't protect AI systems you haven't assessed.
A structured review of how your organisation integrates and operates AI — covering threat models, prompt handling, data flows, access controls and the control gaps that matter most. Not a compliance checklist. A practical security assessment.
Typical focus areas
- AI integration threat modelling — how AI touches your data and processes
- Prompt injection risk assessment — direct and indirect attack paths
- Data inputs, outputs and leakage vectors
- Authentication and authorisation for AI APIs and endpoints
- Privilege boundaries for AI agents and automated workflows
- Third-party AI supply chain and vendor model risk
- Audit logging and monitoring of AI interactions
- Privacy and data handling compliance for AI-processed information
- Output validation and safety controls
What you receive
- AI risk register with severity ratings
- Control gap report mapped to OWASP LLM Top 10
- Priority remediation action plan
- Architecture review notes and improvement recommendations
- Executive summary for board and leadership
- Follow-up briefing session
Data exposure paths
We trace what data enters and leaves your AI systems — including what the model can access, what it retains, and what could leak through generated outputs or API responses.
Access and privilege
We assess whether AI agents and integrations operate under least-privilege, whether API access is appropriately controlled, and whether agentic actions are bounded correctly.
Organisations integrating LLMs or AI APIs into products or internal tools, deploying agentic workflows, using AI-enabled SaaS, or needing assurance over AI risk exposure for governance or compliance purposes.
2–4 weeks depending on the number of AI integrations and complexity of data flows. Simpler single-model deployments can often be assessed in 2 weeks.
Ready to start?
Book a briefing to discuss your AI security posture.
We'll give you a straight read on what's involved, what it costs, and whether it makes sense for your situation.