Successful AI adoption in defence and national security will depend on whether organisations can maintain human skills, judgement and oversight alongside increasingly capable systems, according to a new briefing paper from the Alan Turing Institute's Centre for Emerging Technology and Security (CETaS).
Published on 9 September, the paper is the first output from a wider CETaS project examining intelligence analysis and the future workforce. It draws on evidence from the legal and health sectors, where the paper found AI is already widely used for tasks such as drafting, document review, clinical documentation and diagnosis, but flagged concerns that junior professionals and practitioners could lose opportunities to develop judgement and reasoning skills through routine work. The paper identifies four priorities for organisations adopting AI: building skills through meaningful training, making deliberate choices about which tasks to automate, designing systems that support effective human oversight, and continuously evaluating AI's effect on performance.
For intelligence analysis specifically, the paper said AI can support high-volume tasks such as translation, summarisation and data mapping, but that analysts will still need to assess reliability, interpret uncertainty, recognise gaps in information and make difficult judgements in adversarial contexts.
Ardi Janjeva, Senior Research Associate at CETaS, said AI had "enormous potential to strengthen national security and resilience", but only if human judgement needed to challenge its outputs was preserved, given the stakes of working in an adversarial environment where mistakes can be exploited.
