The Ministry of Defence (MoD) has started using artificial intelligence to automatically classify and secure sensitive documents in an effort to prevent the kind of leaks that have embarrassed the department in recent years.
According to The Times, the new system embeds restrictions directly into files, stopping users from printing, copying, or emailing them without authorization, even if those files leave secure networks.
The rollout comes after several damaging breaches, including the exposure of Afghan applicants’ data under the Relocations and Assistance Policy in 2022.
In that case, personal details of thousands of people including individuals linked to UK intelligence and military operations were mistakenly revealed in an email spreadsheet.
Officials say the new AI system is designed to make sure such human errors cannot be repeated.
The technology is being supplied by Castlepoint Systems, an Australian company whose AI can read, interpret, and apply labels to documents and emails in real time.
UKAuthority reported that the system, branded as “explainable AI,” not only classifies new files but also audits older records to catch mistakes in security labeling.
MoD insiders argue this will reduce reliance on staff who often make errors when manually handling digital evidence and correspondence.
Critics warn the new measures may replace one risk with another. Cybersecurity experts told Financial Times that while automation plugs some gaps, it also creates fresh vulnerabilities.
If hackers discover how the classification AI makes decisions, they could craft documents that trick the system into assigning weak security controls.
Others fear it could block legitimate users from accessing vital files. Analysts also argue that overreliance on automation could reduce accountability inside the ministry.
If a document is misclassified by AI, tracing responsibility becomes difficult, raising the risk that officials may shift blame to the system.
The MoD has defended the project, saying the rollout will be phased and subject to human oversight.
A spokesperson insisted that automation is necessary as the volume of sensitive information continues to grow.
However, tenders reveal that the department is investing heavily in digital automation, including a £180 million contract for AI and edge computing services over the next four years, as reported by DataCenter Dynamics.
For now, the system is still in controlled testing. But if fully implemented, it could mark a turning point in how Britain protects its sensitive information.
Supporters see it as an essential safeguard against future leaks. Detractors say it could give algorithms too much authority in national security.
