Payment buys the work, not the answer.

01Compensation rules

No personnel may receive compensation that is contingent on favorable evaluation outcomes, recommendations, scores, rankings, certification decisions, procurement outcomes, or client adoption decisions. Advisory and consulting fees are scoped as flat, project, or milestone fees; they are not success fees tied to favorable findings.

02Permitted paid services

Fide AI may charge for evaluation work, advisory services, implementation planning, product-lab prototypes, pilots, sponsorship packages, workshops, and research support. Payment buys time, access, analysis, implementation support, or sponsorship recognition; it does not buy favorable findings, endorsement, certification, procurement support, or suppressed criticism.

03Research, evaluation, advisory, and product separation

Fide AI may perform different kinds of work for different audiences. Public research, independent evaluations, private advisory work, and product-lab prototyping must be labeled clearly so clients and readers understand whether Fide AI is measuring, advising, building, or sponsoring a project.

04Conflict of interest identification

Personnel and reviewers must disclose financial relationships, employment relationships, board relationships, commercial interests, and prior public advocacy involving evaluated systems, advised systems, clients, or prospective clients. Disclosures are reviewed before assignment.

05Conflict handling

Personnel with material conflicts are recused from scoring, adjudication, evaluation design decisions, recommendation language, advisory conclusions, and public claims language for affected clients or participants. Recusal is recorded in the relevant client or public deliverable and disclosed when public claims are made.

06Funder and client non-interference

Funders, clients, sponsors, and contributors may not direct evaluation outcomes, recommendations, participant selection, publication timing, claims language, report framing, or suppression of unfavorable findings. Client confidentiality can govern private artifacts, but it cannot alter conclusions.

07Access and scope limits

Evaluation claims must match the access Fide AI actually had. If system access, logs, prompts, retrieval sources, user interface behavior, or deployment context are unavailable or constrained, public and private deliverables should say so and narrow conclusions accordingly.

08Analytic autonomy

Fide AI must retain independent control over test design, analysis, interpretation, caveats, and recommendation language. Participants may correct factual errors or protect confidential information, but they may not require favorable conclusions or remove material limitations.

09Related-entity enhanced scrutiny

Clients or participants with financial relationships to Fide AI personnel, reviewers, contributors, or funders receive enhanced review, additional disclosure, and non-conflicted signoff before findings, recommendations, or public claims are finalized.

10Product-lab evaluation boundary

If Fide AI evaluates a system, prototype, harness, or deployment that Fide AI helped design, implement, fund, or harden, that involvement must be disclosed. Public claims may require independent reviewer signoff, separate scoring roles, or a statement that the work is a collaborative product-lab assessment rather than an independent third-party evaluation.

11Private work and public claims

Private advisory engagements may remain confidential, but any public claim using Fide AI's name must preserve scope, caveats, conditions, and conflict disclosures. No client may imply endorsement, certification, or blanket approval unless that claim is explicitly supported by a public Fide AI deliverable.

12Public disclosure package

When Fide AI publishes an evaluation artifact, it should identify the evaluated system or class of systems, methodology version, evidence sources, calibration status, material assumptions and worldviews, relevant conflicts, funding or sponsorship context, and claim limits unless disclosure would create privacy, security, or pastoral-care risk.

13Operational traces and monitoring data

Work involving production logs, agent traces, employee activity, user interactions, or other monitoring data must define the purpose, minimum necessary data, access controls, retention, deletion, and any permitted secondary research use. Monitoring access does not authorize unrelated surveillance or imply access to private chain-of-thought. Confidentiality and privacy limits must narrow the evaluation claim when they restrict available evidence.

14Human authority and care boundaries

Fide AI does not prescribe whether churches, ministries, schools, families, or builders should deploy AI. Fide AI provides evidence, risk analysis, and design guidance so responsible human decision-makers can judge alignment with their own mission, doctrine, governance, risk tolerance, and pastoral or parental responsibilities.

15Parent and minor safety

Work involving children or family-facing AI should define age assumptions, parental visibility, data handling, disclosure expectations, companionship and dependency risks, tutoring boundaries, and escalation expectations for self-harm, abuse, grooming, medical, legal, or counseling-adjacent concerns.

16Policy updates

Policy changes are versioned and dated. Changes that weaken independence protections require formal governance approval and public disclosure of the rationale.