legalproof.dev
Document · Charter v0.1 · Draft

legalproof.dev Charter

Status: Draft · Version: 0.1 · Last updated: 2026-05-11

1.Mission

legalproof.dev develops open standards, schemas, benchmarks, and certification tests for verifiable legal reasoning systems.

2.Problem statement

Legal AI systems generate fluent prose with no stable, machine-verifiable proof artifact. Conclusions cannot be audited, replayed, or compared across systems. Without a shared protocol, institutional adoption stalls.

3.Core principles

  • Open standards over proprietary moats.
  • Proof over prose.
  • Defeasible reasoning is first-class.
  • Status-at-time is required, not optional.
  • Replayability is the unit of trust.
  • Human review is the design target — not autonomy.

4.ProofGraph standard (LR-001)

See LR-001. The canonical object that links claims, facts, authorities, defeaters, and a typed conclusion.

5.Legal Reasoning Protocol

A thin-waist API: /v1/reason, /v1/verify, /v1/explain,/v1/proofs/{id}, /v1/attest, /v1/sources/resolve.

6.Conformance testing

Public suites for citation replay, temporal drift, jurisdiction swap, defeater handling, and prompt stability. See /benchmarks.

7.Certification program

A six-level ladder (L0–L5). See /certification.

8.Governance model

Independent board, standards working groups, court & regulator advisory council, and an open-source reference implementation. See /governance.

9.Public-interest commitments

Standards are free to implement. Conformance suites are public. Access-to-justice use cases receive first-class support.

10.Roadmap

v0.1 charter and six initial standards (Q2 2026) → conformance v1 (Q3) → first certified implementations (Q4) → court/regulator advisory pilots.
"Any legal AI system can generate text. A LegalProof-compatible system must generate proof."