With #DSEIUK25 just two weeks away, one theme demands the full attention of the defence community, and to be honest the wider industry with the evolution of AI that has moved from a tactical advantage to a fundamental layer of our national security apparatus. It's no longer just a disruptor, it's a defender, a risk, and a responsibility.
The central theme of DSEI UK 2025 is directly aligned with the UK's latest Strategic Defence Review and the anticipated 2030 Integrated Force Structure.
Across complex digital transformation programmes, the fusion of IT, OT, and AI is reshaping operational resilience. We are building smarter, faster systems, from autonomous platforms to predictive logistics. But the core question remains: are we building safer ones?
The uncomfortable truth is that AI hasn't introduced new threats, it has supercharged the old ones. What once required a team of human attackers can now be automated by a single malicious algorithm. The battlefield isn't just physical, it's algorithmic. This is a direct challenge to mission assurance.
The MOD's "Cyber Resilience Strategy for Defence" aims to make Defence a "hard target" in cyberspace, and it is something wider industry needs to adopt.
AI as a Defender: Automating Cyber Resilience
AI is now a critical layer of our defensive cyber architecture. It offers unparalleled speed and scale in protecting our digital and kinetic assets.
In the cyber domain, AI-driven platforms can analyse vast streams of data from command networks to deployed sensors in real-time to detect anomalous behaviour and predict attacks. This allows us to move from a reactive posture to a proactive defence, anticipating threats before they materialize.
In the physical domain, AI enhances operational resilience. Predictive maintenance algorithms can identify potential failures in military hardware from a strike fighter's avionics to a frigate's propulsion system allowing for pre-emptive repairs and minimizing downtime. This ensures our most critical assets are always mission ready.
AI as a Risk: The Algorithmic Attack Surface
However, in my experience guiding projects, I know that AI is a double-edged sword. Its complexity introduces new vulnerabilities that state-sponsored actors and sophisticated adversaries are already exploiting.
The effectiveness of an AI system depends entirely on the data it's trained on. Malicious actors can poison the data used for training, subtly introducing biases or flaws that could cause the AI to fail or make incorrect decisions during a critical moment. Imagine an AI targeting system that has been compromised to misidentify targets, with catastrophic consequences.
The real battlefield isn't just about targeting physical infrastructure; it's about targeting the algorithms that run them. An adversary could launch a sophisticated attack designed to confuse, bypass, or corrupt an AI's logic, rendering it useless or, worse, causing it to act against our interests.
AI as a Responsibility: Governance & Secure AI by Design
This is where my team at Methods focus on;
How can we build trust in a system that can make a decision in a fraction of a second? The "black box" problem, where we don't know how an AI arrived at its conclusion, is a significant ethical and operational challenge. We must demand transparency and explainability in our AI systems.
As the NIST Cyber Security Framework and the CIS Controls remind us (https://www.cisecurity.org/insights/white-papers/an-introduction-to-artificial-intelligence), the fundamentals haven't changed. Our best defence, even in AI-enabled environments still relies on patching, segmentation, least privilege, and secure configurations.
DSEI 2025 let's move beyond the hype and challenge vendors, policymakers, and operators to think beyond capability and consider consequence. Because in the theatre of modern conflict, the most advanced system is only as strong as its weakest algorithm.