Cyber × AI
Security for AI
The AI you build and deploy is now part of your attack surface. This flagship programme teaches your teams to secure it end to end — protecting models and data, defending against prompt injection and adversarial attacks, hardening the AI supply chain, and deploying AI systems securely in line with ISO/IEC 42001.
Enquire about trainingOverview
The moment you build, fine-tune, or deploy an AI system, you have created something that can be attacked — and attacked in ways your existing controls were never designed to catch. Prompt injection turns helpful assistants into liabilities, poisoned data corrupts models from the inside, and a leaky AI supply chain can compromise systems you assumed were yours alone. AI has quietly become one of the most exposed parts of the modern enterprise.
Security for AI is our flagship programme at the intersection of cyber and AI, built for the teams who must secure these systems in practice. We bridge classic security discipline with the threats unique to AI — adversarial inputs, model extraction, data poisoning, and supply-chain risk — and put your people in labs where they both attack and defend realistic AI systems.
The emphasis is on secure-by-design: isolation, least privilege, robust guardrails, rigorous testing, and monitored deployment that maps cleanly to ISO/IEC 42001. Teams leave able to threat-model the AI they build and run, harden it against real attacks, and give the business the confidence to deploy AI without inheriting unmanaged risk.
Who it's for
- Security engineers and architects responsible for AI-enabled systems
- ML and AI engineers building models and applications that must be secure
- Platform and DevSecOps teams operating AI in production
- CISOs and security leaders extending their threat model to AI
- Risk and assurance professionals evaluating AI system security
What's covered
- The AI threat landscape — how attacks on AI differ from traditional ones
- Prompt injection, jailbreaks, and adversarial inputs against LLM systems
- Model and data security — poisoning, extraction, inversion, and theft
- Securing the AI supply chain — datasets, foundation models, and dependencies
- Secure AI architecture — isolation, least privilege, and guardrails
- Protecting sensitive and personal data across training and inference
- Red-teaming and testing AI systems for security weaknesses
- Secure deployment and monitoring aligned with ISO/IEC 42001
Format & delivery
- Instructor-led, hands-on, delivered on-site or live virtual
- Labs attacking and defending realistic AI systems
- Cohort sizes tuned to your team — typically 6 to 16
- Multi-day delivery scoped to your stack and risk profile
- Tailored to your models, platforms, and deployment patterns on request
Outcomes
- Teams that can threat-model and harden AI systems they build or operate
- Practical defences against prompt injection and adversarial attacks
- A secure-by-design approach to AI deployment across the lifecycle
- Security practices that map cleanly to ISO/IEC 42001 expectations
Industry relevance
Frequently asked questions
How is securing AI different from securing any other software?
AI introduces failure modes traditional security never had to consider — prompt injection, training-data poisoning, model extraction, and adversarial inputs among them. The programme bridges classic security practice with these AI-specific threats so your teams can defend the whole system.
Do participants need a machine-learning background?
A working understanding of how AI systems are built and deployed helps, but the programme is designed for security professionals as well as AI engineers. We level-set the AI concepts so security specialists can engage fully.
Is there hands-on attacking and defending?
Yes. Participants work in labs that both exploit and harden realistic AI systems, because the most effective defenders are those who understand how these systems are actually attacked.
How does this relate to AI governance?
Security and governance are complementary. Governance sets the policy and risk framework; this programme delivers the technical controls that make secure AI real. It pairs naturally with our AI Governance programme.
Can the labs use our own AI stack?
Where you share your models, platforms, and deployment patterns in advance, we tailor the exercises so the defences transfer directly to your environment.
Download the datasheet
Get the full programme outline, delivery options, and example agenda as a PDF.
Ready to train your team?
Tell us about your team and we'll recommend the right courses and curriculum.
Talk to us