FIFTH DOMAIN’s work is based on a simple idea: people and systems should be understood through what they can demonstrate, not what is assumed about them.
The company was founded in Canberra in 2015 by Matt Wilcox. Its early work focused on practical cyber learning and assessment, giving people a way to build capability through realistic cyber activities. This reflected a problem we saw in the market: qualifications, interviews and self-assessments could not always show what someone would actually do in a practical task. Evidence could.
As AI systems have started to take on more cyber work, the same problem has expanded. Organisations now need to understand not only what people can do, but how AI agents behave, where they fail, what oversight they need and whether they are suitable for the mission.
FIFTH DOMAIN builds technology, evaluation environments and intelligence systems that help examine these questions. Our work shows how cyber operators and AI agents act, decide and perform under realistic conditions, so organisations have better evidence before relying on them.
We refer to this as Cyber Force Intelligence: evidence from cyber missions that helps organisations understand human and AI capability, improve development and support accountable decisions.
FIFTH DOMAIN is owned by Australians and operated by Australians in Australia.
What we work on
FIFTH DOMAIN’s work is concerned with how cyber capability is demonstrated, observed and understood. We aim to make the evidence clearer, so the people responsible for decisions can make them with greater confidence.
A large part of this work is focused on human cyber capability. We support practical learning, assessment and development through realistic tasks that show how people investigate, make decisions, use support, recover from errors and develop over time.
As AI agents become part of cyber work, we are applying the same evidence-led approach to understanding their behaviour. This includes examining what an agent does, how it uses tools, where it succeeds, where it fails and what evidence is available to support decisions about its use.
Much of this work depends on the design of the mission itself. FIFTH DOMAIN builds mission environments, objectives and conditions that allow specific behaviours and capabilities to be observed. The results can then be connected and examined to understand gaps, patterns and next steps.
FIFTH DOMAIN’s Cyber AI research extends this work by examining AI-agent behaviour and the evaluation methods needed to understand it. This work produces cyber evaluation data, training data and behavioural insights for model builders, AI developers and research partners.
