A 501(c)(3) Research Institute — Baltimore, MD

Trauma-informed governance for AI systems that carry consequence.

Decisions about safety, escalation, and care now move through AI systems that were not designed for the people they act on. TIAIL translates the clinical framework of trauma-informed care into system requirements, evaluation protocols, and deployment criteria, then tests them in the settings where the stakes are highest.

Core framework

Trauma-informed AI governance

Systems that cause harm without accountability repeat harm. TIAIL builds governance standards that center the experience of those most exposed to algorithmic decision-making, from clinical settings to public infrastructure.

Research pillars

Three pillars

Pillar 01

AI regulation & policy

Institutional review protocols, deployment criteria, procurement frameworks, and policy language for public systems.

Pillar 02

Trauma-informed evaluation

Evaluation rubrics, audit protocols, and model benchmarks calibrated to clinical and human services settings.

Pillar 03

Public toolkits and compliance infrastructure

Open-source tools and documentation frameworks for Medicaid, HUD, and grant compliance, freely available to nonprofits and public agencies.

Framework

The SAMHSA six principles, operationalized as system constraints.

A thirty-year clinical framework for structuring care environments so that disclosure is possible, applied dimension by dimension as pre-deployment criteria for AI in health and social services.

Applications

Where this work lands.

Trauma-informed AI design applies across every system that serves people navigating crisis, recovery, or survival. Six active application areas:

This is the phase in which categories are defined, and TIAIL operates within this narrow window of influence.

The clinical literature on trauma-informed care is precise, cumulative, and underused in the design of systems that routinely act on the people it describes. The gap between that literature and the design of AI in health and social services is the size of the harm those systems are already producing.

Closing that gap is the work.

Active research

Field-embedded. Iterative. Cumulative.

Programs move from framework formulation → first-pass artifact → field validation → publication. First-pass artifacts are released for public review; at-scale deployment follows empirical outcome data, because the populations these systems act on cannot afford to be the validation set.

Founding charter

AI is the fulcrum. Trauma-informed systems are the mission.

Every research question, every model built, every governance framework published traces back to this singular purpose: ensuring that artificial intelligence does not replicate or amplify the harms experienced by those it is meant to serve.

TIAIL is

TIAIL is not

We build. We test. We validate. We influence.

Our outputs include computational ontologies, evaluation rubrics, audit protocols, and model benchmarks — not position papers. The emphasis is on the artifact, the evidence, and the effect on policy.

News & updates

News & updates.

Milestones in the work, with external announcements and cross-links.

2026-05-19

Update

Cohort delivers first-pass crisis support framework.

A four-person Spring 2026 cohort at the MedStar-Georgetown AI CoLab delivered a first-pass trauma-informed AI crisis support framework for domestic violence survivors. Built over three months under TIAIL sponsorship, with privacy and safety carried as primary design constraints from initial scoping rather than retrofitted. The prototype is public for review. A second cohort will continue the iteration.

Prototype repo Announcement

2026-05-18

Update

Talk: building AI to carry consequence.

AIM-AHEAD Consortium hosted Session 3 of its Distinguished Speaker Series on May 18, 2026: "Building AI to Carry Consequence: The Case for Trauma-Informed AI across Systems of Care." The talk maps SAMHSA's six trauma-informed principles to system requirements for AI in regulated care settings.

Announcement

Get involved

Get involved.

Decisions about safety, escalation, and care are running through AI systems most people will never see. Anyone working on those systems can test, audit, and improve them with TIAIL's infrastructure: models, evaluation rubrics, and testing frameworks, public and free to use. The work is public by design.

Volunteer Newsletter Discord GitHub Donate

Research methodology

Research methodology.

Frameworks intended for at-scale deployment are validated against empirical outcome data prior to release. First-pass prototypes are released earlier, for public review and iteration. Outputs span computational ontologies, evaluation rubrics, audit protocols, and model benchmarks.