Overview
The UK government's AI Cyber Security Code of Practice and NCSC guidance on secure AI development both stress testing AI systems before and after deployment. We test chatbots, RAG pipelines and AI agents for prompt injection, data leakage and unsafe actions.
Generative AI features change your threat model. An attacker no longer needs a code bug if they can persuade your model to ignore its instructions, reveal another user's documents, or call a tool on their behalf.
Our AI security assessments focus on your application — prompts, retrieval, tools, permissions and output handling — rather than the underlying foundation model. We test direct and indirect prompt injection, data leakage across users and tenants, excessive agency in agents, and unsafe handling of model output in downstream systems.
Beyond testing, we help you set AI usage policies, assess third-party AI tools and align with emerging frameworks such as the NIST AI RMF, ISO/IEC 42001 and the EU AI Act. Because most AI features sit on ordinary APIs, we usually combine AI-specific testing with a web and API assessment.
How testing works for UK organisations
- Signed scope and written authorisation before any testing, in line with the Computer Misuse Act 1990
- Testing windows agreed in UK time, including out-of-hours windows for production systems
- Fully remote delivery, with internal and wireless access via VPN or a posted test device, so there are no travel costs
- Reports mapped to ISO 27001, PCI DSS, UK GDPR Article 32 and customer questionnaires, with a fixed price in GBP
What's included
Our approach
- Architecture reviewModel provider, orchestration, retrieval, tools and data flows.
- Threat modellingMapping abuse cases to the OWASP LLM Top 10 and MITRE ATLAS.
- Adversarial testingManual and assisted prompt attacks across direct and indirect channels.
- Integration testingWeb/API testing of the surrounding application and tool endpoints.
- Report & guardrail designPrioritised fixes and secure-design recommendations.
What you receive
- Findings mapped to OWASP LLM Top 10 (2025)
- Prompt injection test corpus results
- Data isolation verification
- Agent permission review
- Secure AI design recommendations
Standards & frameworks
- OWASP Top 10 for LLM Applications 2025
- MITRE ATLAS
- NIST AI RMF
- ISO/IEC 42001
- UK AI Cyber Security Code of Practice
- NCSC secure AI development guidelines
Frequently asked questions
Can prompt injection be fully prevented?
Not reliably with current models. The practical goal is to limit impact: treat model input as untrusted, restrict tool permissions and enforce authorisation outside the model.
Do you test the AI model provider?
No. We test your application's use of the model. Model providers operate their own security programs under their terms.
Is this a separate engagement from a web app test?
It can be, but we usually combine them because most AI features are delivered through ordinary web and API endpoints.
Can you support UK organisations from India?
Yes. All services are delivered remotely by our team in India. India is 4.5 hours ahead of UK time during British Summer Time and 5.5 hours ahead in winter, so our working day overlaps with yours every morning and afternoon. Testing windows and escalation contacts are agreed around your UK schedule, and 24/7 services run around the clock.
Where is our data processed?
Engagement data is handled by our team in India under NDA, with encrypted storage, least-privilege access and secure deletion at the end of the agreed retention period. We can sign a data processing agreement using the UK International Data Transfer Agreement (IDTA) or the UK Addendum to the EU standard contractual clauses where personal data is involved.