Mind Foundry: Trusted AI for Distributed Sensor Intelligence
Ahead of APEX Europe 7-8 October 2026, we had a Q&A with Mind Foundry, a UK sovereign AI company spun out from Oxford University’s Machine Learning Research Group. They spoke about sensor fusion, acoustic intelligence, trust in AI, and where the next opportunities lie for UK defence.
As defence organisations face an unprecedented growth in sensor data across every domain, the challenge is no longer collecting information, but turning it into timely, trusted intelligence. For Mind Foundry that challenge sits at the heart of its focus.
Mind Foundry came out of Oxford University research in Bayesian methods and signal processing. Do you feel that the connection between academia, industry and the military is closer than it has been for a long time, and what does it mean practically for defence customers?
The pace of change in both technology and the threat environment means defence can no longer rely on innovation moving through long, linear development cycles.
The closer the relationship between operators, researchers and developers, the faster ideas can move from research into operationally relevant outcomes. Defence provides the operational problem, academia contributes scientific advances and new ways of thinking, and industry focuses on turning those ideas into deployable capability. Our connection with the University of Oxford helps us maintain access to leading expertise while remaining focused on solving real operational challenges.
Practically, that means defence customers gain access to capability that evolves more quickly, can be refined in response to user feedback, and can benefit from advances in research without waiting for entirely new hardware programmes.
You have two main products, SENTRY for sensor fusion and NIGHTINGALE for anti-submarine warfare. Are these in operational use, or still in development and trials? Can you say who with?
Both products have been developed and evaluated alongside defence customers and partners in operationally relevant environments:
NIGHTINGALE has been integrated and assessed through collaborations involving the UK Ministry of Defence, the Royal Navy, Thales and SEA. It has demonstrated the ability to support near real-time acoustic classification and tonal analysis, helping operators reduce workload and accelerate decision-making. The capability has progressed beyond evaluation into operational deployment and is now supporting a funded defence programme.
Similarly, SENTRY has been demonstrated in operationally relevant scenarios spanning counter-UAS and underwater applications, including participation in UK-US defence experimentation activities.
NIGHTINGALE tackles the acoustic analysis challenge in ASW, a domain where the data volumes and complexity are enormous. What does your AI bring that current approaches don’t?
One of the most important trends in anti-submarine warfare is the growth in sensing.
Future ASW concepts increasingly depend on networks of crewed platforms, uncrewed systems, fixed sensors and distributed acoustic arrays. The result is a dramatic increase in the amount of data available to operators.
The challenge is that more data does not automatically produce better decisions. NIGHTINGALE is designed to help bridge that gap. It automates large parts of the tonal analysis workflow, including denoising, tonal detection, signature characterisation, support and contact prioritisation. Instead of presenting operators with raw data, it highlights what matters and provides evidence-based outputs to support assessment.
We often talk about machine-speed decision support rather than machine-made decisions. The aim is not to replace warfare specialists but to reduce their cognitive burden and accelerate understanding, so operators can focus their expertise where it adds the most value.
SENTRY promises a single fused operational picture from multiple sensors. What sensor types does it currently integrate, and what platforms is it deployed on or designed for?
SENTRY was developed to address a challenge that has become increasingly important across every domain: the inability of humans to manually correlate growing volumes of data from multiple sensors, alongside the increasing number of threats we need to defend against. It’s becoming more complicated to build a complete picture.
SENTRY, therefore, is designed to fuse information across sensors, specialising in acoustics. Rather than treating each data source independently, SENTRY combines them into a coherent and continuously updated operational picture.
It’s also built for environments where sensors are imperfect. Different sensors update at different rates, experience different types of degradation and perform differently under varying conditions – SENTRY accounts for those realities rather than assuming every sensor behaves the same way.
We see this kind of sensor fusion becoming increasingly important as defence moves towards more distributed and networked sensing architectures across multiple domains.
Trust and transparency in AI are a big conversation in not just defence but most industries right now. How do you make your models accountable enough for commanders to trust them in high-stakes decisions?
Trust is not something that can be added at the end of development – it needs to be designed into systems from the beginning.
One of the risks in today’s AI conversation is focusing too heavily on model performance without paying enough attention to assurance and operational integration. In defence, those factors are often more important than achieving marginal improvements in accuracy. Our approach combines explainable outputs, robust testing, model provenance, post-deployment monitoring and human oversight. We put significant emphasis on understanding how models were developed, what data was used, how they are performing and whether performance is changing over time.
We also believe assurance should be an enabler rather than a blocker. The goal is not to slow adoption unnecessarily, but to provide the evidence commanders need to understand both the strengths and limitations of a system. Defence is ultimately built around accountability. Human operators and commanders remain responsible for decisions, and AI should help them make those decisions with greater confidence and awareness.
What are you planning to show and hope to discuss at the inaugural APEX Europe event in October?
APEX Europe is about solving real operational problems, so we’ll be focusing on how AI can enable operators make sense of an increasingly complex sensor environment.
We’ll be showcasing our work in across sensor intelligence and fusion, discussing the importance of trustworthy AI in defence, and exploring how technologies such as distributed sensing, autonomy and decision support can help deliver decision advantage.
Most importantly, we’re looking forward to hearing directly from operators and defence leaders about the challenges they need industry to solve next.
Where do you see the biggest near-term opportunity for AI in UK defence, and where is the MoD still too cautious?
The biggest opportunity is helping operators cope with complexity. Across maritime, air, land and national security environments, the volume of sensing information available to operators continues to grow. AI can help fuse, filter and prioritise that information, allowing humans to focus on judgement rather than data processing. We believe that is where some of the most immediate and practical benefits will be realised.
Defence organisations are understandably careful when introducing AI into mission-critical environments – the consequences of failure in defence are too significant for reckless deployment, and rigorous assurance and testing are vital. If there’s an area where defence can continue to evolve, it’s the speed at which assured AI capabilities move from experimentation into operational adoption. Once technologies have demonstrated clear value and appropriate assurance, there’s an opportunity to accelerate deployment and iterative improvement. Recent conflicts have demonstrated how quickly software-enabled capabilities can adapt, and defence organisations that can safely harness that agility are likely to gain a significant advantage.
Find out more at https://www.mindfoundry.ai/
