How Singapore’s SAFER Unit Flags Child Protection Risks Earlier

Singapore’s Ministry of Social and Family Development (MSF) launched its Sensemaking Alerts for Frontline Engagement and Response (SAFER) unit on September 22, 2026. It is intended to help child-protection professionals see concerns that might be harder to recognize when each agency holds only part of a family’s information.2

How SAFER identifies cases for review

SAFER starts by bringing together relevant information from MSF, the Singapore Police Force and the Central Narcotics Bureau. A risk algorithm analyses the available information to surface cases that may warrant closer attention. Its initial focus is families with children under seven.2611

An algorithmic flag is not a determination that abuse has occurred. Child-protection professionals assess the information alongside what they know about the family and decide whether a safety check, further assessment or support is needed. The purpose is to inform and speed up human judgment, not replace it.26

Why Megan Khung’s case prompted the change

Four-year-old Megan Khung died in February 2020. A subsequent review identified failures in how concerns about her safety were communicated, recorded and followed up across the agencies involved. SAFER is part of MSF’s effort to make those separate warning signs more visible to practitioners.41021

MSF has said that, had SAFER existed at the time, Megan’s case would have been flagged for review about five months before her death. That is a retrospective assessment of when a review might have begun—not proof that an alert would have prevented her death.19

What its first results show—and what remains unclear

In its first two months of operation, SAFER surfaced about 600 cases for further review, with roughly 70 leading to alerts or closer follow-up. Those figures describe the screening and review process; they do not mean 70 children were confirmed to have been abused.614

MSF plans to extend SAFER to more agencies and age groups.5 Professional review remains the central safeguard described in the available accounts. Those accounts do not establish the precise limits on data access or details of privacy and algorithm-audit procedures, so the early figures alone cannot show how accurately the system identifies risk—or how well those safeguards work in practice.26