Raytheon UK takes lead in AI-driven defence training solutions

The AI-powered training tool, developed by integrating MAInt technology from UK-based small firm ScotAI, offers a robust system for efficiently searching and managing technical training content for complex military platforms that have vast numbers of document pages.

“The key advantage of MAInt lies in its ability to use AI to efficiently extract relevant content from extensive databases, following pedagogical frameworks like Bloom’s taxonomy,” said Alex, chief technology officer for Raytheon UK’s Training Transformation Services.

It’s not just about sorting data; it’s about asking the right questions, Alex said. Unlike many AI tools that rely on cloud-based systems, Raytheon UK’s solution is a secure, on-premises system designed to meet the Ministry of Defence’s stringent requirements that call for tailored, secure solutions that meet specific operational needs.

Developing the AI tool involved rigorous security controls in close collaboration with the customer, data management, and integration with Raytheon’s own systems to meet defence standards, such as JSP 740 Acceptable Use Policy for Information and Communications Technology and Services.

“This was crucial for maintaining the integrity and confidentiality of sensitive training materials. By integrating Bloom’s Taxonomy within the JSP 822 Defence Systems Approach to Training, it helps us use the right verbs at the right time and level,” said Alex. “For example, if a training objective is to operate a diesel generator, you can request, ‘Provide 500 words on operating the diesel generator.’ The AI, utilising the pedagogical framework, will generate the content. 

“You still need a learning specialist to review and refine the content, but the system indicates where the content resides, thus easing the workload. Our future goal is to integrate this output into the right templates to enhance team efficiencies and provide relevant technology and capability to our customers,” Alex said.

AI assurance and risk mitigation

When it came to AI assurance, security controls, and risk mitigation, Raytheon UK drew on expertise from within its Data Capability Group – which houses cybersecurity, data engineering, and AI/ML specialist teams – to properly implement and optimise the AI system for customer requirements.

Led by Head of Data Science & Machine Learning, Katie, the SRG (Strategic Research Group) focuses on AI research and has been tasked with providing AI assurance and consultancy during the system roll-out, and works with the software development team to address any issues.

“We’re providing risk mitigation, AI assurance and consultancy as the tool is rolled out. This involves evaluating the capabilities of the tool from an AI perspective and working through practical levels to assess and mitigate any risks of using the AI tool, ensuring accuracy and usefulness of outputs.”

Katie emphasised the strategic priority of identifying and mitigating technical AI risks, such as “model hallucination,” where AI tools might provide inaccurate outputs. “It’s all about keeping up to date and making sure that we’re delivering the best quality training to the customer with the most efficiency in order to produce reliable, high-quality training material faster.” 

Data security and threat management

The implementation and security controls around the AI tool involved the cybersecurity team. The team conducted a threat modelling exercise to identify key threat actors and risks, and then implemented various security measures such as on-premises hosting, data storage and processing controls, user restrictions, penetration testing, security monitoring, and secure destruction of data.

The AI tool sits within Raytheon UK. Key challenges have included integrating data from various sources currently sitting within MOD systems and ensuring the security and integrity of that data as it is aggregated and processed by the AI tool.

This security and AI assurance work is essential and emphasises the comprehensive approach to the security of sensitive data and addressing vulnerabilities.

“Our threat assessment helps us to look at what the appropriate security controls are, mitigate risks and provide confidence in the tool’s use,” said James, head of Cyber Security. “It’s strategically important to embed security from the start of projects, along with continuous threat modelling exercises, security monitoring and secure destruction practices, at the end of engagements, to protect data integrity. Supply chain infiltration, for instance, is a major concern and ensuring that we’re not impacted by third-party vulnerabilities.”

A vision for the future

The next generation of the tool could see human factors integrated into AI-driven training environments. Alex said he and his colleagues are exploring using the tool and associated AI and machine learning capabilities for other applications like biometric monitoring of trainees and for improving training content and design.

“We know that human factors influence combat decision-making. By incorporating biometric data into our AI-driven training tools and analysing mass – such as eye tracking, voice modulation, and physiological responses in simulators – our goal is to create a dynamic learning dashboard enabling real-time monitoring and post-training analysis to identify learning gaps and optimise training outcomes,” said Alex.

Raytheon UK’s tool not only supports the MOD’s training initiatives but also sets a new standard for AI application in the defence sector, promising a future where training standards, technology and defence strategies are seamlessly integrated.

“We plan to use the technology across multiple mission areas to drive improved quality, efficiencies and SME engagement,” Alex said.