More data, same deadline: can AI help fix the SAR bottleneck?
Public services have always been data-heavy. Long before digital transformation became a familiar phrase, teams were recording information about the people they support, from handwritten case notes and assessments, to detailed accounts of some of the most sensitive moments in people's lives.
Today, much of that information is stored digitally, alongside emails, photographs, audio and video. The formats may have changed, but the purpose remains the same - supporting people, delivering services and recording decisions.
Behind every record is a person. Someone receiving care, a child needing support, or a family navigating difficult circumstances. And when someone exercises their right to access that information, protecting the privacy of everyone involved can become a complex task.
More information, growing pressure
Subject Access Requests (SARs) can involve reviewing years of records, identifying information about other individuals and making careful decisions about what can lawfully be disclosed.
For social care services, this can be particularly challenging. Records may contain deeply personal accounts, family relationships, safeguarding concerns and information about multiple individuals.
The pressure is already visible. In September 2026, a Borough Council reported that SAR volumes had risen by 45%, from 162 in 2020/21 to 235 in 2024/25. Despite improvements in performance, approximately 100 requests remained overdue, predominantly complex social care cases requiring substantial document review and redaction.
And this council is not alone. Across public services and other data-heavy sectors, organisations face similar pressures. Growing volumes of information, historic records and repetitive manual processes place increasing demands on teams. Under pressure, these processes also create opportunities for human error.
The challenge is clear - more information to review, stretched resources and statutory deadlines that remain unchanged.
Where can AI help?
AI offers an opportunity to remove much of the repetitive groundwork, identifying potentially sensitive information across documents, images, audio and video.
Rather than spending hours manually searching through records, reviewers can start with information already identified for their attention. They can check the findings, add or remove proposed redactions and make decisions based on the circumstances of each request.
The original records remain intact. AI provides the head start, people remain responsible for the outcome.
Responsible adoption matters
Introducing AI into processes involving sensitive personal information rightly raises questions about accuracy, privacy and accountability. But using AI responsibly shouldn't mean handing over control.
The College of Policing's guidance on AI identifies automated redaction as a potential starting point for organisations beginning their AI journey. It also emphasises the importance of choosing technology that addresses a genuine operational need, with appropriate safeguards and human oversight.
These principles extend well beyond policing. Technology providers should work with organisations to understand their governance requirements, demonstrate safeguards and support testing, transparency and auditability. AI should fit within an organisation's accountability framework, not compromise it.
Ultimately, modernising these processes isn't simply about working faster. It's about protecting sensitive information while ensuring people can exercise their right to access it. Responsible AI shouldn't change who makes the decisions, it should give the person who does, more time to make the right ones.
