Advisory and applied research in
privacy-preserving AI.
Equitable Labs advises institutions in finance and health care on responsible
AI: setting the rules, testing systems against them, and running the research that shows what
holds up inside a real institution.
Every engagement starts with a problem you bring us. We examine it with the
people who own it, work out what the evidence supports, and leave you with answers your team
can run and defend without us. When a problem needs shared infrastructure, Goose FL, our
technology practice, builds it.
Equitable Labs is accepting partner
requests in finance and health care.
We are a research and advisory practice and a public
benefit corporation: we exist to make AI answer to the people it is hardest on, wherever it
is used. If you need to decide whether an AI system is safe to adopt, or show that one works,
tell us the problem.
Equitable Labs is a research and advisory practice. Goose FL is its technology practice, and
builds the privacy-preserving AI infrastructure the work needs.
01 · What we do
Four ways we work with institutions: we advise, test, teach
and build.
Strategic AI advisory
Equitable Labs advises leadership teams on the decisions that outlast a procurement
cycle: which systems to adopt, what evidence a board needs before it approves one, and who
is accountable when a model makes a call about a person.
The work runs from assessing vendors and models to mapping governance onto the rules an
institution already answers to, such as the EU AI Act, ISO 42001 and the NIST AI Risk
Management Framework. Most of it is ongoing, because the questions do not stop when a
project does.
Applied research and pilots
Equitable Labs designs studies and runs pilots that answer a question no single
institution can answer alone: whether an AI agent is safe to put into a live system,
whether a credit model excludes the people it should serve, whether hospitals can learn
from one another's records without moving them. Each study is documented well enough that
another institution can act on the result. Methods include federated learning, secure
computation and homomorphic encryption.
Where a pilot puts an AI agent into a live system, it runs under NOMOS, the governance
layer Goose FL builds for agents. NOMOS sits between the agent and the systems it acts on,
and checks every action against the institution's rules before it happens. Each decision
is recorded for review.
Equitable Labs teaches boards, executive committees and the risk, compliance and clinical
informatics teams that run the work how to understand, govern, evaluate and question an AI
system: AI literacy, lifecycle governance, impact assessment, model evaluation and
algorithmic ethics. Boards come away with the questions to ask before they approve a
system, and the standard of evidence they will need to defend the decision.
Team members have lectured at:
Cornell University
Milwaukee School of Engineering
University of Maryland Global Campus
Jönköping University, Sweden
MKE Tech, Milwaukee
Meltwater Entrepreneurial School of Technology (MEST), Accra, Ghana
Technology
Sometimes a problem needs shared infrastructure, so we build it. Goose FL is the
technology practice of Equitable Labs.
It lets several institutions work on one question at once while each keeps its records
where they are. What comes out is a model, a measure or a finding, not a copy of anyone's
data. It runs inside each institution's own environment, under the approvals that
institution already operates by.
Status: Goose FL holds a memorandum of understanding with the National Information
Technology Development Agency, and has submitted a cross-institution fraud-detection use
case to the second cohort of the Central Bank of Nigeria's regulatory sandbox (as at
November 2026).
Ongoing advisory for a network of institutions across several regulatory
environments.
Most of our advisory work is ongoing. In one relationship we advise the coordinating
office of an institutional network and, through it, the member institutions it supports
across several regulatory environments.
Tool and vendor assessmentWe map what is available and what is credible, then
test the shortlist against fairness and inclusion measures rather than against a
vendor's demonstration.
Fairness and ethics reviewEvery tool, vendor and proposed use case is reviewed
for the exclusion it might cause, and each review leaves a document that a funder or a
regulator can follow.
Governance and regulatory alignmentWe turn the rules members already answer to,
from banking and health care law to the EU AI Act, ISO 42001 and the NIST framework, into
controls a team can run and an examiner can check.
Frameworks we map the work to
GDPR
EU AI Act
ISO/IEC 27001
ISO/IEC 42001
NIST AI RMF
HIPAA
ForHumanity Fellows Katie and Paul help draft its certification
schemes and lead modules of its AI Policy Accelerator, which builds a company's
compliance controls against those schemes.
03 · Applied research and pilots
Studies and pilots, run with the institutions that need the
answer.
We run studies and pilots with institutions that need an answer they cannot get alone.
The work spans health care studies that bring the analysis to patient records in place,
financial inclusion research, AI governance and policy frameworks, and compliance testing
inside regulatory sandboxes, all aimed at the equitable use of AI in the public
interest.
A financial inclusion project in NigeriaIn progress · with the National IT Development Agency,
the Presidential Commission on Economic and Financial Inclusion (Office of the Vice
President) and a local governance expert NGO
A privacy-preserving oncology study of patient records across
hospital networksUnder review · Submitted to the National Institutes of
Health, now awaiting assignment to a scientific review group
A six-stage bank review that tests credit models for
inclusionUnder review · Submitted to ACM FAccT 2027, for bank
model risk teams
Some studies need a research partner before they can be funded.
We join grant and consortium proposals as the named research partner, write the method,
and set what the pilot has to produce, so the funder gets evidence rather than
activity.
Tell us
Tell us what is not working.
Two or three sentences is normally enough. We answer within two
working days, and we will tell you plainly if it is not a fit.