Research→Programs→Projects→Publications→News→People→Join→Contact us
Independent research institute

Building trustworthy AI for public safety.

Public Safety AI Lab develops scientific methods for evaluating, verifying, governing, and deploying AI in settings where reliability is not optional.

Explore our researchFour research programs

Public safety is the hardest test of AI. A mistake is not a bad recommendation — it is a wrongful outcome. PSAIL builds the evaluation, governance, and assurance science that high-stakes deployment demands.

0

Research programs

0+

Active research tracks

0M

Synthetic personas

0+

Papers in peer review

Research Programs

Four programs, one mandate.

We organize the lab around research programs — not papers. Every publication, patent, and project connects back to one of these four.

All programs→
01

AI Evaluation

Rigorous, domain-grounded benchmarks that measure whether AI systems are reliable enough for public safety.

PoliceBenchLLM-as-JudgeBenchmarkingReliability
02

AI Governance

Institutional methods for deploying AI under due process, accountability, and procedural justice.

Procedural JusticeAccountabilityResponsible AIPolicy
03

AI Assurance

Verification and evidence pipelines that make an AI decision auditable, traceable, and defensible.

VerificationProvenanceEvidenceFidelity
04

AI Systems

Agentic systems and simulation environments that turn research into deployable public-safety workflows.

Agent SystemsSimulationWorkflowDeployment
Signature Feature

The Research Map.

Every research asset is a node in one connected graph. Trace a policy outcome back through the project, the patent, the paper, and the program that produced it — the full provenance of an idea.

ProgramPaperPatentProjectDataset & CodePolicy Impact
Explore the map→
PROGRAMAI GovernancePAPERProcedural Justice as DesignPATENTControl-Layer MethodPROJECTPublic Safety AIDATASETGovernance ToolkitIMPACTInstitutional Adoption
Selected Work

Research that ships.

From benchmarks to deployed systems — our projects turn evaluation science into tools institutions can actually use.

All projects→

PoliceBench

Research

AI Evaluation

An evaluation benchmark for language models on real policing tasks — task corpora, difficulty grading, and human-expert baselines for measuring where models are, and are not, reliable enough to assist.

GABM

Active

AI Systems

A simulation platform that populates policy scenarios with LLM-driven synthetic personas — built on a million-scale Korean population — to stress-test interventions before they reach the public. Applied to fraud prevention, disinformation response, and more.

1MKorean synthetic personas

Public Safety AI

Research

AI Governance

A governance framework for adopting AI inside public-safety institutions — third-party verification of capability and procedural-justice controls that map oversight, accountability, and redress into concrete operating rules.

모두의경찰관

Research

AI Systems

A citizen-facing police AI platform studied as a case of AI organizational augmentation — how source, trust, and adaptation shape adoption among citizens and officers (master's research).

Newsroom

What's new.

Releases, publications, talks, and milestones from the lab.

All news→
Announcement

Voice-phishing defense GABM submitted to the Journal of Police Policy

A generative agent-based study of voice-phishing defense strategies for older adults enters peer review.

Talk

EV-Hub edge AI presented at the KCC 2026 Urban AI workshop

Work on dynamic DNN partitioning for public-safety edge networks was presented at KCC 2026.

Announcement

모두의경찰관 research advisory board launched

An advisory board convened for the master's research on AI organizational augmentation in policing.

Publication

EV-Hub edge AI paper accepted at JKICS

The camera-ready for the EV-Hub split-computing paper was submitted; publication forthcoming.

Join the lab

Help build the science of trustworthy AI.

We work with researchers, engineers, and public institutions who take reliability seriously. There is real work to do.

Work with us→Get in touch