Work

Strengthening Child-Protection Reporting with Privacy-First AI

Year: 2026
Type: article
Abstract: How an on-premises AI intake assistant can support U.S. hotlines, Children's Advocacy Centers and investigators. U.S. child-protection professionals handle an extraordinary volume of reports: an estimated 4.4 million referrals to child protective services in federal fiscal year 2024, and 21.3 million reports to the NCMEC CyberTipline. Every report deserves to reach a trained person quickly, completely and in the right order. This paper proposes a privacy-first AI intake assistant that works alongside existing hotlines, online portals, Children's Advocacy Centers and investigators. It gives reporters a simple, always-available way to make a complete report by text or voice; it gives intake staff structured reports, clear summaries and urgency flags with stated reasons; and it runs entirely on agency infrastructure, so no child's data leaves the agency. The paper shows how each common type of report would be handled, what each group of users gains, and how the approach can be tested safely, from synthetic data to a supervised pilot. Trained professionals make every protective decision, and each state configures the assistant to its own laws and chosen level of automation; the assistant helps them spend their time where it matters most. Independent research. This paper is not affiliated with or endorsed by any employer or government agency, uses only publicly available information, and is not legal advice.
Language: English
Open Access status: green
OpenAlex ID: W7225871078
FWCI:
Cites:
Cited by:
Related to: 10