What we believe about AI shapes what we build.

This is the conviction behind our work: AI will not create equitable outcomes on its own. We have to advise, advocate and build for them.

Front page of the Equitable Labs commitment: a person on a ridge at sunrise, October 2026.

Read it in ten minutes.

Ten pages: the bargain, six things we hold, and how the co-founders came together.

Document
The Equitable Labs Commitment
Edition
First edition
Format
A4 PDF, 10 pages, 2.1 MB
Open the Commitment (PDF)

Benefit does not arrive on its own.

Every generation is offered a technology that will fix something, and asked to give something up in return. AI arrives with a particular bargain: if you want the benefit, hand over the data that describes your people, your patients and your borrowers, and trust whoever holds it to use it well.

Most of the world is being asked to accept that bargain without seeing the terms.

We are not against the technology. We use it every day, and some of the most careful people we know are building it. What we object to is the assumption that benefit arrives on its own, and that the institutions serving most of the world's people will catch up later. They will not catch up later. Capability compounds in the places that already have it, and the distance shows up in decisions taken without the people those decisions land on.

What follows is what we hold, written down so that we can be held to it.

Two men bargaining across a market stall, one gesturing with open hands, the other reaching back, with colored awnings above them.

Six statements we can be held to.

  1. Start with the person the system lands on

    Before we ask what a system does, we ask who it does it to. A decision that lands on a person is a moment in somebody's week: a loan declined, a claim flagged, a patient triaged. Those moments are where risk becomes real, and they are the least represented in the rooms where systems are designed. An assessment that cannot name the person on the receiving end of the decision is not finished.

  2. Evidence, not assertion

    This field runs on confidence. We prefer proof. A governance program that cannot show what it measured, on what evidence, and what happened when the answer was unwelcome is a document rather than a control. So we build the artifact as well as the argument: the test, the record, the method somebody else can run without us in the room.

  3. Intelligence can be shared without the data moving

    The data that matters most is the data that cannot move: patient records, loan books, case files, the things an institution holds on behalf of people who never agreed to anything. Learning together has always seemed to require pooling, and pooling is exactly what the people the data describes cannot consent to. It does not have to be that way, and we work on methods that let institutions improve a shared model while each keeps custody of what it holds.

  4. The capability gap is the equity gap

    Adoption is not a matter of enthusiasm. It takes people, money and time that a small regulator, a municipal government, a cooperative or a rural clinic does not have. Equity in AI adoption is therefore a distribution of capacity: the ability to specify what you need, to test what you are sold, to run it afterward, and to refuse it. We work on the capacity side, with the institutions that have the least of it and the most to gain.

  5. Harm is local, and so is the evidence

    Risk looks different in Apia and Kigali than it does in Brussels. A model that is safe against one benchmark can be useless in a language with few speakers, or dangerous where a family shares one phone. Global frameworks are necessary and not sufficient. The people who can tell you how a system fails are usually the ones living with the failure, and they are rarely asked.

  6. Say what we are for, in public, and report on it

    We are a Delaware public benefit corporation. That is a legal duty before it is a badge: our charter names the benefit we exist to produce, and our board has a duty to weigh it against return. We will report on what we did about it, including the parts that did not work.

Milwaukee's river at blue hour, the downtown buildings reflected in the water.

None of us set out to start this company.

Katie and Paul met on the same piece of work: drafting the certification schemes that providers and deployers of high-risk AI systems are measured against at ForHumanity. Writing an audit criterion is a specific discipline. You cannot write one without deciding what evidence would satisfy it, and that decision is where a regulation either becomes a control or stays a paragraph.

Ed joined when the work turned to the hardest question in the EU AI Act. With a consortium of thirteen authors we wrote our bid for the European Commission's tender on modeling and evaluating sociotechnical risk in general-purpose AI: how a model's behavior combines with the context it is placed in, and how you would assess that before the harm rather than after it.

Lindsey came to the work from the public sector, where she had seen how easily a strong policy can stall between adoption and implementation. Her focus was on what it takes to make governance usable: bringing the right people into the room, giving them clear responsibilities, and helping institutions act on what they learn.

Then the work moved. Through the UNDP's AI Trust and Safety Re-imagination Programme we all worked on AI safety with teams in countries where the assumptions behind European and American frameworks do not hold. In Samoa and in Rwanda, and in the other places the program took us, the questions were different and more concrete. Whose language does the model not speak? What does safety mean where a household shares one phone? None of it could be answered from Brussels, and most of it had already been answered locally by people nobody had asked.

We came back with the conviction the practice is built on. The gap in AI is not a gap in ambition or in talent. It is a gap in capacity: the ability to specify a system, to test it, to refuse it, and to keep the evidence afterward. The institutions with the least of that capacity serve the most people, and they are the ones being asked to accept terms they had no part in writing.

So we started Equitable Labs as a public benefit corporation, with a mandate that describes the benefit we exist to produce. Alongside it we started Goose FL, its technology practice, to build the privacy-preserving AI infrastructure that lets institutions learn from one another without moving their records.

Another advisory practice on its own would not change much. One that can write the standard, test against it, and build the thing it recommends might.

Two colleagues in conversation at a round table, a printed sheet and coffee cups between them.

Four of us, four different ways into the same problem.

Katie Grillaert

Katie Grillaert

  • AI Research
  • Algorithm Ethics
  • AI Adoption
  • Workshops & Lectures

Katie helped draft ForHumanity's EU AI Act certification schemes for providers and deployers, leads a regulatory sandbox on Article 9 (risk management) of the EU AI Act, and is a certified auditor of AI and algorithmic systems. She advises a project at UNRWA, the United Nations agency for Palestine refugees, on the ethics of general-purpose AI in humanitarian settings, and publishes on how risk-averse financial algorithms shut people out.

Paul Crafer

Paul Crafer

  • AI Product Management
  • AI Governance
  • Business Case Assessment
  • Workshops & Lectures

Paul came to AI governance after twenty years leading technology teams. He drafts EU AI Act certification schemes with ForHumanity, keeps a working liaison with the European Data Protection Board, the Austrian and Dutch data protection authorities and Swedac, the Swedish accreditation body, and works on AI policy and governance projects with organizations such as GIZ, the German development agency, and ACET, the African Center for Economic Transformation. The technology background is what he brings to a use case: what a system is for, who it lands on, and what a regulator would need to see before trusting it.

Ed Vocke

Ed Vocke

  • AI Systems Design & Development
  • Data Science
  • Technology Research
  • Workshops & Lectures

Ed has spent a decade turning frontier AI research into systems that survive contact with a regulator. He co-authored 'Applying Privacy-Enhancing Technologies to LLMs in Critical Infrastructure Contexts' (IEEE CARS 2025). He builds audit infrastructure and secure deployments for institutions that cannot ship their data anywhere, and works on the measurement problem underneath: what a model's behavior means once it is in front of a loan officer or a clinician.

Lindsey Washburn

Lindsey Washburn

  • AI Governance
  • Public Benefit
  • AI for Public & Social Sector
  • Stakeholder Engagement
  • Workshops & Lectures

Lindsey came to the work from the public sector, where she saw how easily a strong policy stalls between adoption and implementation. She led enterprise AI governance engagements with senior government officials, designed a secure and responsible technology lab, and ran an enterprise-wide AI governance program for a US city government. Her focus is what makes governance usable: the right people in the room, clear responsibilities, and institutions able to act on what they learn.

Tell us what the problem is.