Governance Infrastructure for Legal Reasoning

From Black Box to Crystal Box

Every decision auditable. Every reasoning chain reproducible.

We don't build AI that guesses legal answers. We build infrastructure where legal expertise becomes deterministic code, auditable logic, and reproducible proof.

Deployed with Parlamento Federal del Clima Tajamar Foundation

Black Box to Crystal Box — The Legal AI Crisis

Your LLM gives you a legal answer. You ask: how did you reach that conclusion? It can't tell you. It's not broken — it's probabilistic, that's what neural networks do. But in a regulated market, where every decision must be defensible, auditable, and compliant, probabilistic is a liability.

Real Costs of Black-Box Legal AI

Case 1 — The Interpretation Problem

We used AI to interpret EUDR compliance for our supply chain. Six months into export, the auditor rejected our reasoning. We had no documented justification — just "the model said so." Result: $1.2M in renegotiated contracts.

Case 2 — The Contradiction Problem

Our legal AI recommended contract terms without flagging that provincial labor law contradicted national tariff policy. The contradiction existed in our own data — the AI just missed it. We found out during litigation.

Case 3 — The Reproducibility Problem

We're scaling to 5 new provinces, each with different environmental regulations. Do we have a documented, auditable method for ingesting local laws? No. Each jurisdiction is manual. Each decision is invisible. Each audit is a surprise.

The Fragmentation Crisis: Why LatAm Is Bleeding

15,000+

Argentina — environmental laws across 24 jurisdictions with zero interoperability.

LexML

Brazil — the standard exists, but no unified governance layer sits on top of it.

Federal → State → Municipal

Mexico — regulatory cascade creates combinatorial chaos across levels.

3-4% GDP

Entire region — lost annually to regulatory friction (Inter-American Development Bank).

The region doesn't need more laws. It needs a system that finally understands the laws it has.

The Regulatory Acceleration

EU AI ActEffective August 2027 — explainability and auditability mandatory for high-risk systems.
EUDRLive December 2025 — supply chain traceability for $11.85B of Argentine exports.
CSDDDCorporate due diligence decisions must be documented and reproducible.
National AI strategies$500M–2B allocated for legal digitalization across LatAm.

These aren't distant policy ideas. They're active today, and your compliance proof must exist before audit finds gaps.

Crystal Box: what LLMs can't do alone.

Ulpianus AI is not a better chatbot. It's not RAG over case law. It's a governance infrastructure that separates reasoning logic from linguistic output. Most legal AI: LLM reads cases → makes inferences → decides → database stores data. Ulpianus AI: expert legal reasoning → coded as deterministic governance rules → database executes the rules → decision is computed → LLM narrates the output. The edges are deterministic. The output is linguistic.

The Architecture: Three Layers

01

Expert Curation (You Own This)

Your legal experts read provincial ordinances, tribunal rulings, contradictory regulatory frameworks and temporal changes in legislation — then model the logical structure: conditions, exceptions, conflicts, and when rules changed. This becomes your Semantic Matrix — pure logic, no LLM, no vectors, no embeddings.

02

Deterministic Governance Engine

A query enters in natural language. The engine routes it to the correct legal domains, queries your Semantic Matrix deterministically, checks for contradictions (fail-closed if ambiguity exists), computes the answer with certainty — not probability — and generates a proof chain. Same query → same answer → same hash. Always. This is the MSM-IAR layer (Multidimensional Semantic Matrix — Inferencia, Activación y Razonamiento).

03

Proof Chain (Auditability Built In)

Every decision is signed with SHA-256 and timestamped in the PoI Ledger (Proof of Inference). You can prove when a decision was made, how it was computed, and why it was correct according to the rules your experts defined. Auditors see the full chain. Regulators see reproducibility. You own the proof.

The Trino Brain: Three Reasoning Modules

CUCO

Epistemic Integrity

Cognitive Uncertainty and Confidence Orchestrator. For every inference, CUCO generates a confidence vector: source credibility, logical consistency, bias detection, completeness. It ensures the system never hallucinates — it knows what it doesn't know.

LUTHOR

Logical Reasoning

Logical Understanding Through Heuristic Orchestrated Reasoning. Performs dialectical analysis: Thesis (extract the normative claim), Antithesis (identify contradictory regulations), Synthesis (harmonized interpretation with legal grounding).

CUTHOR

Executive Integration

Consensus Understanding Through Hybrid Orchestrated Reasoning. Merges CUCO and LUTHOR outputs into compliance certificates, legislative impact assessments, and regulatory gap analyses — decision-ready intelligence.

Why This Beats Everything Else

Competitor ClassProblem
Generic LLM + Legal DataDoes not handle contradiction. Does not audit. Does not prove compliance. Hallucinates jurisprudence.
Case Law RAGFinds relevant cases. Doesn't reason from first principles. Doesn't scale to local ordinances. No auditability.
Consultant-Driven SaaSOpaque reasoning. Vendor lock-in. You don't own your legal models. Auditors can't reproduce decisions.
Watson Orchestrate (IBM)Externalizes governance to third-party runtime. Vendor controls reasoning. Not sovereign.

Ulpianus: your experts define the logic. The database governs the reasoning. You own the proof. Auditors see the chain. Regulators see reproducibility. Zero vendor dependency.

The nine-dimensional reasoning space.

MSM-IAR operates across 9 independent dimensions of legal reasoning. Law doesn't live in 1D — EUDR compliance interacts with local tariff policy, which interacts with labor contracts, which interacts with environmental zoning. The 9-dimensional space handles n-dimensional reasoning without combinatorial explosion.

DimWhat It TracksWhy It Matters
D1Semantic MeaningWhat does this law actually say?
D2Epistemic CertaintyHow sure are we of this interpretation?
D3Ethical AlignmentIs this interpretation ethically sound?
D4Temporal ContextWas this law in effect on this date?
D5Jurisdictional ScopeDoes this law apply in Jujuy but not Córdoba?
D6Actor AttributionWho issued this rule? Are they authorized?
D7Relational TopologyDoes this rule depend on, or contradict, another?
D8Traceability VectorProof-of-Inference ledger entry.
D9Physical IntegrationReal-world compliance verification (IoT, sensors, custody chain).

LLM-Agnostic Orchestration

Critical innovation: Ulpianus sits above the LLM layer, not within it. LLMs (GPT-4, Claude, Gemini, LLaMA) serve as linguistic translators — converting semantic vectors into human-readable text. The reasoning logic resides in the metacognitive orchestration layer. This is not a preference. It's a structural moat.

01

No Vendor Lock-in

Swap foundation models without architectural redesign.

02

Cost Optimization

Route queries to the most cost-effective model that satisfies accuracy.

03

Composable Intelligence

Use specialized models per task within one workflow.

04

Future-Proof

As LLMs evolve, Ulpianus adapts without retraining the core system.

05

Compliance by Design

External orchestration enables end-to-end auditability.

The Stack

Data Layer

  • PostgreSQL + TimescaleDB
  • Bitemporal modeling
  • PostGIS (jurisdictional reasoning)
  • pgvector (semantic integration)

Governance Layer

  • CUCO Protocol
  • LUTHOR Engine
  • CUTHOR Synthesizer
  • AKN4AR standard

Proof Layer

  • PoI Ledger (SHA-256 signed)
  • Time-travel capability
  • Immutable audit chain

Translation Layer

  • LLM orchestration (any model)
  • SSE streaming
  • Natural language interface

Not concepts. Auditable infrastructure in operation.

Active pilot · 2026

Digesto AI

Parlamento Federal del Clima + Tajamar

  • 91 provincial laws ingested from Chaco, converted to AKN4AR
  • Expert-curated by Rodrigo Galarza and team
  • Deterministically queryable across environmental law domains
  • Full PoI Ledger audit chain

Proves: multi-jurisdiction ingestion is reproducible, AKN4AR works at scale, curators can model tacit expertise. Foundation for national-scale Digesto Ambiental.

Completed 2026

SP-BOS

Retail Legacy Migration

  • 72 tables migrated from MySQL kiosk system to PostgreSQL
  • 10,000+ rows of transactional data
  • Zero data loss
  • All governance logic deterministic

Proves: SETNET's DBOS philosophy works for legacy systems with 100% data integrity.

Production operational

PRIME

TYC Industrial Vertical

  • 32 migrations on real manufacturing data
  • WORM PoI Ledger operational
  • Time-travel queries working
  • Dimensional veto matrix (D1-D6) preventing bad decisions

Proves: the architecture works in capital-intensive, high-risk industrial contexts, at real-world complexity.

IP & Competitive Moat

  • Patent pending: "Cognitive Architecture for Hybrid Semantic Reasoning Systems" (USA, EU, Argentina jurisdictions)
  • Exclusive ownership: Ulpianus retains all datasets, algorithms, and outputs from pilot projects
  • First-mover advantage: only provider with triple interoperability (AKN4AR / LexML / AKN4EU)
  • Barrier to entry: USD 300K+ investment required to replicate the foundation

A realistic, adjusted timeline.

Timeline reflects realistic operational velocity. Initial phases compressed vs. early roadmap due to focus prioritization and institutional validation sequencing.

Phase 0Q4 2026 – Q1 2027In progress

Technical Validation & Reproducibility Proof

Establish scientific credibility and reproducibility.

  • Formalize 20-query benchmark suite with peer-reviewed methodology
  • Complete OCR and legal parsing optimization for the Argentine corpus
  • Publish reproducible Proof of Interpretation (PoI) whitepaper
  • Patent prosecution active
Phase 1Q2 – Q4 2027Pilot active, scaling

Digesto Ambiental Deployment

Demonstrate production-grade deployment in government context.

  • Deploy unified environmental digest across 24 Argentine jurisdictions
  • Launch "Clima y Ley" public portal (citizen-facing legal search)
  • Onboard Chaco, Córdoba, CABA as pilot jurisdictions
  • Full audit trail + PoI Ledger for all queries
Phase 2Full year 2027–2028Commercial

Legal App & Commercial SaaS

Commercialize to private legal sector and expand institutional users.

  • Release Legal App v1.0 for law firms and in-house legal teams
  • Launch Cursor Legal (parliamentary intelligence subscription)
  • Onboard 5–10 law firm pilot customers
  • Expand Digesto to Chile, Uruguay, Paraguay
Phase 32028+International

EUDR Compliance & International Expansion

Establish Ulpianus as the regional RegTech standard for ESG/compliance.

  • EUDR supply chain traceability module
  • ESG compliance vertical
  • University partnerships (10–15 institutions)
  • Market entry: Brazil, Chile, Colombia

Unit economics overview.

Segment Performance (Annual Metrics)

SegmentAvg Contract ValueGross MarginPaybackLTV/CAC
Government$165K60–70%5.0 mo29.8x
Judicial$130K55–65%5.4 mo22.3x
Corporate$47.5K70–80%2.7 mo29.7x
Legal Firms$18K75–85%2.9 mo20.6x
Academic$10K70–80%4.0 mo25.0x

All segments exceed the 3:1 LTV/CAC benchmark considered healthy for SaaS.

Revenue Scenarios 2025–2030

Base Case Organic Growth

$3.34M $29.86M

EBITDA 48.7% (2025) → 67.3% (2030) · Break-even 2025 Q1

Accelerated Strategic Partnerships

$5.33M $70.35M

EBITDA 57.6% (2025) → 70.0% (2030) · Break-even 2025 Q1

Conservative Government-Focused

$2.00M $17.07M

EBITDA 32.5% (2025) → 63.5% (2030) · Break-even 2025 Q1

All scenarios cash-flow positive from Year 1.

Key Financial Metrics

$120M → $578.7M

Valuation range, Jan 2026 → Dec 2028 base case.

Optional

Series A needed for acceleration, not survival — company self-funds on base case.

-20% → +40%

EBITDA trajectory, 2026 → 2028.

$3.7M

Non-dilutive funding projected from grants (BID, CAF, EU programs).

Ready to build Crystal Box legal infrastructure?

For Technical Founders & Angels

12-page deep dive on MSM-IAR architecture and reproducibility guarantees.

Download Technical Brief →

For Institutional Partners

Digesto expansion, EUDR compliance, regional deployment.

Partnership Inquiry →

For Law Firms & Corporates

See Cursor Legal and compliance automation in action.

Book Demo →

For Regulators & Government

Legal AI infrastructure design.

Institutional Integration →
hello@ulpianus.ai

From Black Box to Crystal Box™

Artificial intelligence will transform legal systems. The question isn't if — it's how. Will that transformation be opaque and unaccountable? Or transparent, deterministic, and verifiable? Ulpianus AI chooses transparency. We convert legal expertise into auditable code. We make reasoning reproducible. We build infrastructure where governance lives in the edges, not the nodes.

This is not a feature. This is architecture.

The regulatory window is open for 18–24 months. The first provider to achieve scale in deterministic legal reasoning will set the LatAm standard for compliance infrastructure. We're building that provider.