Legal and compliance work depends on source authority, jurisdiction, version history, citations, evidence, and human review. AI in this domain cannot be a generic chatbot. An answer that can’t be traced to authority is a liability, not a feature.
How OEP fits
- Citations are the product. Retrieval is evidence-verified. Every answer is checked against the verbatim source text it claims to rest on, and insufficient evidence is a first-class verdict, surfaced rather than papered over.
- Version-aware legal navigation. Supersession chains and an evidence-anchored citation graph (cites, amends, repeals, exceptions) answer “is this still the law?” structurally.
- Authority-aware ranking. Instrument level, court level, and recency are normalized metadata, so ranking is deterministic and explainable across packs.
- Refusal as governance. A platform-level refusal taxonomy distinguishes researching the law from seeking advice, enforced below the app where policy can’t be skinned away.
- Human review is structural. Packs ship with their review status on their face. Jurisdiction packs are ratified by licensed lawyers before they’re labeled ready. The pipeline doesn’t let unreviewed content masquerade.
What exists today
This is OEP’s most developed vertical: populated, evidence-verified statute packs across multiple jurisdictions, a drafting-template contract with bilingual templates under legal review, and LegalVecta, the consumer expression of the same foundations, in active development. See LegalVecta →
What we won’t tell you
LegalVecta and OEP’s legal foundations do not give legal advice, do not substitute for lawyers, and are not ready for every jurisdiction today. Human legal review remains necessary by design, not as a disclaimer. See how we bound our claims.