Security Capability Development
for People & AI
Training, detection and response exercises and hacking competitions for people; tailored evaluation and improvement of AI agents
Learn, try it for real, compete, and grow.
We build the practical capability of people and AI to carry out security work and to detect and respond to attacks. The judgements people make, the actions they take and the feedback they get carry into their next piece of work.
Can our people and AI agents make sound judgements and get real security work done?
People build capability through education, hands-on exercises and hacking competitions, and AI agents through evaluation and improvement.
Expert reviewed
Training programmes
Four programmes for people, by role and purpose. Evaluating and improving AI agents is covered separately below.
Can our team look at our systems the way an attacker would?
Security staff and developers learn the attacker's view first-hand in a controlled lab, from penetration testing, bug hunting and reversing to embedded, industrial and AI systems.
When a real attack comes, do our detection and our people catch it and respond?
Detection and response staff train against real attack techniques alongside the attack side, one technique at a time, and fix what was missed on the spot.
How far have our people's skills come, and who inside or outside the organisation stands out?
We run internal, industry and public competitions, from challenge design to running the event: participants test their skills against each other, and you find the people who stand out.
When a suspicious message arrives, do our staff stop to doubt it and report it?
Through realistic situations, such as phishing emails or impersonation calls, all staff build the habit of doubting first and reporting straight away.
Designed for each programme
Goals, length and conditions are set together for each programme.
Learning goals
What participants should be able to judge and do afterwards, tied to their role and work.
Length and format
From a single workshop to a longer programme that links several courses by role, set to your goals and schedule.
Participants and conditions
Who takes part, at what level, and conditions such as the practice environment.
Run training alongside testing or validation, and the weaknesses they uncover can become the subject of the programme itself.
AI agent evaluation and improvement
A tailored engagement that evaluates and improves how well an AI agent does security work and how safely it operates. It is shaped around your agents and the work you give them: agreed tasks, and an evaluation set that can be run again.
- Task performance
On tasks close to the real work, we evaluate whether the agent plans towards its goal, uses knowledge and tools to carry out the security work it is given, and checks its own results. The tasks and the criteria are set with you, and the set is built so the same evaluation can be run again after changes.
- Safe operation
We check that operating controls such as permissions, approvals, interruption and recovery hold in practice, and that people can stay involved throughout the work and step in when needed.
- Improvement
Gaps found in evaluation feed improvements to the knowledge, tools, procedures and review flow the agent works with. You keep the results by task, the improvement proposals and the evaluation set itself. Where new tools or agents are needed, we can build them with you through technical collaboration.
Here we look at whether AI agents can do security work. Testing an AI system as the target of attack is covered by AI System Security Assessment & Penetration Testing.