01
Evaluate an AI system
Get independent evidence for a model release, deployment, or expansion decision.
Contact and collaboration
Bring us a system or consequential question you need to understand.
We work with AI labs and organizations in high-trust domains to evaluate complete systems and build better ways to measure trustworthiness. Start with a short structured intake so we can understand what is at stake and determine the right next step.
Ways to work together
How can we make AI more trustworthy?
Bring us a system, domain expertise, a research question, or the resources to investigate one seriously.
02
Bring domain expertise
Help define what trustworthy AI requires in a consequential setting you know well.
03
Advance the research
Turn an important open question into evidence that others can inspect and build upon.
04
Support the work
Fund, sponsor, or convene independent research that strengthens the evidence for trustworthy AI.
Developing or deploying agents?
Enterprise teams and implementation partners can explore a scoped design partnership to develop repeatable evaluations for a consequential workflow.
Individual contributors
Want to get involved individually?
The Fide AI Contributor Network is for individuals who want to offer expertise, technical work, review, lived experience, communication, or practical help. Contributors do not need an institutional sponsor or a particular religious commitment. We value people who can state their assumptions, examine evidence, and engage disagreements seriously.
Introduce yourself as a contributor ↗Ways to contribute
01
Research and methods
Study design, construct validity, statistics, literature, and peer review.
02
Engineering and evaluation
Harnesses, datasets, trace analysis, security testing, and reproducible tooling.
03
Domain practice
Firsthand knowledge of consequential decisions, standards, and real workflows.
04
Community perspective
Lived experience, affected-community insight, and missing questions or assumptions.
05
Governance and institutions
Policy, law, privacy, procurement, standards, and organizational oversight.
06
Communication and convening
Writing, editing, education, visual explanation, workshops, and introductions.
Participation may range from one review to a defined project, an ongoing expert pool, open-source work, or a future paid role. Scope, compensation, credit, ownership, and confidentiality are agreed before work begins.
Where trust matters
Every high-trust domain brings different risks and responsibilities.
High-trust domains are settings where people make consequential judgments from context, take high-stakes action balancing risk and reward, shape relationships, and deliver care.
Workplaces and enterprise
Professional judgment, agentic workflows, oversight, and accountability.
Finance
Transactions, financial advice, risk decisions, compliance, and customer trust.
Legal and justice
Legal analysis, professional judgment, due process, rights, and accountable authority.
Cybersecurity and critical systems
Defensive agents, access boundaries, threat analysis, incident investigation, and control of high-consequence operations.
Healthcare and care
Clinical judgment, caregiving, wellbeing, and decisions affecting vulnerable people.
Education, youth, and families
Learning, development, household decisions, and responsibility for children.
Public service and civic life
Rights, public decisions, service delivery, and institutional legitimacy.
Faith and religious life
Source fidelity, spiritual authority, formation, pastoral boundaries, and institutional life.
Open research
Make the research broadly useful and accessible.
We publish research, methods, benchmarks, datasets, and tools so people beyond Fide AI can inspect the evidence and carry the work forward.
Public research repositories
Explore Fide AI's code, datasets, evaluation methods, and reproducible research artifacts.
View Fide AI on GitHub ↗
Open calls for research
Browse defined questions that Fide AI and independent researchers can investigate, critique, or carry forward.
Explore the research-ideas repository ↗
Independence
We take research independence seriously
AI labs, clients, sponsors, reviewers, domain experts, and institutions can participate under published rules without buying favorable findings, preferred timing, endorsement, certification, or product claims.