How Fide AI protects trust.

This library summarizes the policies and methodology that make the public mission inspectable: governance, conflicts, benchmark design, claims discipline, entity separation, funding transparency, and release accountability.

Governance charter

Fide AI separates methodology, evaluation operations, advisory input, publication decisions, and funding activity. The goal is to protect benchmark outcomes from donor, sponsor, participant, or related-entity influence.

Independent decision rights for methodology, scoring, publication, corrections, and appeals.
Documented review for conflicts, recusal, and related-party transactions.
Public accountability through claims limits, funding transparency, and release caveats.

Entity structure

Fide AI is the nonprofit public-standard institution. Product companies are the better fit for Public Benefit Corporation logic: product development, subscriptions, partner deployments, commercial tooling, equity, and investment capital.

Fide AI nonprofit Standards, research, benchmark governance, public reporting, field education, scholarship access, and independent evaluation.

Product-company path Products, subscriptions, commercial faith-AI infrastructure, partner deployments, customer contracts, and employee/investor equity.

Conflicts and related entities

Covered contributors disclose relevant interests, recuse from conflicted decisions, and document recusals. Related participants require disclosure, enhanced oversight, and non-conflicted signoff.

Related entities receive no early access to hidden scenarios or scoring rubrics.
Conflicted personnel do not control evaluation, adjudication, appeals, or publication signoff.
Related participants may cite results only under the same public claims policy as every other participant.

Benchmark and methodology

The benchmark package defines scope, scenario handling, scoring dimensions, run manifests, release versions, limitations, and correction processes. Scores are tied to named benchmark versions and evaluated configurations.

Measured behavior Theological and pastoral quality, grounding, preference fidelity, comparative honesty, and escalation appropriateness.

Known limits Benchmarks cannot replace institutional judgment, pastoral discernment, human review, or real-world monitoring.

Claims policy

Fide AI results are evidence of benchmark behavior. They are not theological authority, pastoral authority, universal product approval, or a guarantee of performance outside the evaluated conditions.

Funding transparency

Fide AI will report funding-source categories, sponsorship constraints, and relevant conflicts so the field can understand who supports the work and how independence is protected.

Donors and sponsors cannot influence rankings, certification decisions, appeals, methodology freeze decisions, or publication signoff.
Funding supports public-good infrastructure, reviewer calibration, field education, and access support.
Commercial work, if any, must be mission-aligned and governed by documented non-interference rules.

Want the full policy packet?

Express interest