Responsible AI framework
VoxJurisPath treats every AI feature as informational preparation, never legal advice. Four safeguards are enforced in the product, not promised in a policy page.
Start free · See how it works
Serving legal aid organizations, courts, and 50-state pro se litigants. Not a law firm. Not legal advice.
The four verified safeguards
Exhibit-ID whitelist. Server-side validation refuses to render any exhibit ID that does not appear in the user's Exhibits & References field. The AI cannot invent an exhibit; unsupported statements are flagged USER-STATED or UNSOURCED.
Citation verification. Case citations are cross-checked against CourtListener and statutory citations against Cornell LII before display. On verification failure the citation is not silently dropped — it is surfaced as unverified.
Delivery-only coaching rubric. The rubric is scoped server-side to clarity, pace, composure, evidence anchoring, and time. It cannot score merit, credibility, or outcome, and the server strips the rubric entirely for non-coaching deck types.
Pre-export placeholder gate and role-play testimony refusal. PDF export is disabled until every bracketed placeholder is resolved, and role-play decks emit [ANSWER — TO BE PROVIDED BY WITNESS] rather than fabricate testimony.
Alignment with external frameworks
Each safeguard maps to ABA Formal Opinion 512's duties of competence, confidentiality, and disclosure for generative AI, and to National Center for State Courts guidance on preserving neutrality in court-facing tools.
Human-in-the-loop and review
Institutional deployments include supervisor review flows, with a supervising-attorney dashboard planned for Q1 2027. Administrative and sensitive AI actions are logged via database triggers so audit records cannot be silently skipped.
Important: VoxJurisPath is not a law firm, does not provide legal advice, and is not a substitute for a licensed attorney. We provide self-help software and legal information.