Research
Research for trustworthy AI.
Fide AI studies AI systems through the lens of making them more trustworthy. We research whether systems use evidence well, remain within legitimate authority, preserve human control, and stay accountable while they act.
This research develops and validates the methods independent assurance depends on. We publish evidence and public-good artifacts that others can inspect, challenge, and use in their own work.
What trust demands
No single discipline can make AI trustworthy.
AI systems cannot become trustworthy through capability testing alone. Trust also depends on how systems use evidence, exercise authority, affect human judgment, operate within institutions, and behave in consequential settings.
These questions cross technical, social, moral, and institutional disciplines. Our research agenda maps their connections and turns them into concrete calls for research. Some will be pursued by Fide AI. Others are invitations for researchers, labs, institutions, and domain experts to take forward.
09
Connected research areas
86
Public calls for research
Open
A research commons built for contribution
Research areas
The questions receiving focused attention.
01
Frontier capabilities
What new risks appear as systems become more capable and autonomous?
02
Agent alignment
Does an agent remain aligned while it plans, uses tools, and acts?
03
Evaluation science
Can the evidence support the decision people want to make?
04
Grounding and truthfulness
Will the system use the right source and represent it faithfully?
05
Human agency and formation
What judgment, habits, and relationships does AI use shape?
06
Faith and religious life
How should AI behave around religious sources, authority, practice, and community?
07
High-trust domains
What does trustworthy behavior require in a consequential setting?
08
Governance and readiness
Which safeguards and authority structures should govern deployment?
09
Work and vocation
How does advanced AI change useful work and human contribution?
Published research
Evidence from our first high-trust domain.
Our released studies currently focus on faith-facing systems. Each page states the question, method, result, artifacts, and limits so readers can inspect the evidence directly.
Featured research
14 frontier models evaluated
When AI Is Your Pastor
Do clearer instructions improve how AI handles theological, moral, and pastoral-adjacent questions?
Open research page →
Paper 01 · FID-056
8,640 matched requests
When Not to Generate
Can AI quote Scripture exactly? We tested four ways of producing a quotation and traced where each one can fail.
Open research page →
Latest paper · Sacred Text Fidelity series
4,800 controlled observations
Knowing When to Defer
Will AI check an authoritative Scripture source when a user asks it to rely on memory?
Open research page →
Current research directions
Questions moving toward public evidence.
A research direction names work in progress. It does not imply that Fide AI has reached a finding.
How do advanced AI systems create cybersecurity risk?
An emerging direction for independent investigation of agent capabilities, authority boundaries, oversight, and complete-system behavior.
Can organizations see when an agent departs from its task or policy?
Developing privacy-aware operational traces, deviation tests, escalation criteria, intervention controls, and recovery measures.
Do evaluation results support the decisions made from them?
Testing construct validity, reviewer reliability, benchmark gaming, and the gap between controlled evaluation and deployed behavior.
Will a system use the right source when pressure makes that inconvenient?
Measuring source delegation, retrieval, citation fidelity, instruction hierarchy, and deterministic delivery designs.
Where should AI defer to responsible people and institutions?
Studying authority boundaries, human handoff, oversight, relational substitution, and the effects of repeated use.
Emerging research direction
Independent research on AI behavior and cybersecurity risk.
Fide AI aims to conduct rigorous, independent research on how advanced AI systems behave in cybersecurity settings. We seek to investigate how agents use their capabilities, cross authority boundaries, respond to oversight, and create risks that become visible only when the complete system is examined.
Our perspective emphasizes legitimate authority, evidence integrity, human agency, and accountability. We make the assumptions behind our research explicit so others can examine both the methods and the conclusions.
Cybersecurity is an emerging research direction for Fide AI. We are building on our published work in source fidelity and system behavior, and seeking security researchers and partners to develop and test methods in this domain.
Interpretation limits
A result describes behavior under named versions, prompts, conditions, rubrics, and evaluation procedures. It does not establish general safety, product approval, or deployment readiness beyond the tested scope. Faith-domain studies also do not confer theological or pastoral authority.