Ethics of AI in Defence Meeting

APPG for Defence Technology – Ethics of AI in Defence

Date: Monday 9th June 2025

Location: Room Q, PCH, Palace of Westminster

Attendees:

Matt Walker (Leonardo), James Clark (APPG Secretariat), Chloe Sanderson (APPG Secretariat), Sarah Moth-Lund Christensen (University of Sheffield), Ken Turley (RUK), Leo Karaalp (Leonardo), Ben Craig (Montfort), Hamish Mundell (RUSI), Ash Risby (Vizgard), Jonathan Bray (University of Sheffield AMRC), Gareth Hetheridge (Leonardo), Kyle Thomas (SAIF Autonomy), Hannah Furse (Office of Fred Thomas MP), Rupert Small (Egregious), Georgina Robinson (Mishcon de Reya), Toby McCrindle (Mishcon de Reya)

Meeting Minutes

 Leonardo Presentation: AI Risk Management and Integration in Defence

 Leonardo provided an overview of its operations in helicopters, cyber, space, and electronics, highlighting the central role of digital engineering and AI across the business.

 Their electronics division particularly relies on software, firmware, and data engineering.

 AI has long been integrated into Leonardo’s products to generate decision advantages in defence applications.

 They have developed a bespoke AI Risk Framework to manage 15 AI use cases across operations, engineering, and in-service teams.

 The framework addresses six categories of risk: legal/regulatory, safety, commercial, data, AI security, and technology.

 AI use is managed at the use case level (e.g., image detection, facial recognition), not just at the component level.

 Leonardo has created an AI Assurance Register to log risk assessments and ensure traceability and accountability.

 The framework aligns with international standards including ISO, NIST, and the upcoming JSP 936 Part 1, while also anticipating the EU AI Act.

 The policy is operationally embedded with clear accountability, overseen by a data subcommittee reporting to group level.

 Emphasis was placed on the need for regulation that supports innovation without unnecessarily stalling progress.

 SAIF Autonomy Presentation: Ethical AI and the Trust Layer

 SAIF Autonomy introduced their “trust layer” concept for autonomous systems to ensure operation within ethical parameters.

 The presentation stressed the importance of embedding ethical considerations into system design, particularly where autonomy is physically enacted.

 Highlighted the need for government funding and regulatory frameworks that enable responsible development and deployment of AI technologies.

 Raised concerns about the challenges of regulating proprietary AI systems, and advocated for focusing on infrastructure and processes rather than the models themselves.

 Discussed the balance between innovation and oversight, and called for more coherent, centralized regulation to avoid fragmentation across the sector.

 Emphasized the importance of collaboration between large defence primes, SMEs, and government in shaping AI policy.

 Underscored the ethical risks associated with AI in military decision-making, especially in targeting and high-stakes scenarios.

 Advocated for robust human-in-the-loop systems, high-quality data management, and improved data-sharing frameworks to support safe AI adoption.