HAREUtil

Every query. Every role. Every document. Proof.

HAREUtil is designed to provide runtime governance for utility AI—from offshore wind planning to grid operations. Role-based access with cryptographic evidence intended to support audit, judicial review, and FOIA workflows.

Sector Disclaimer: Vertical examples illustrate intended application domains. Sector-specific compliance requires independent legal and technical validation.

The Challenge

Utilities Have an AI Trust Problem

You want AI assistants that can read the entire project file—environmental assessments, grid studies, public submissions, internal memos. But regulators, courts, and the public need to know: what did the AI see, and did it respect legal boundaries?

What You Have Today

AI assistants with broad file access
No proof of what informed AI answers
Role-based access that doesn't extend to AI
Judicial review risk if AI touches privileged docs
Former employees with lingering access

What You're Missing

AI that respects role boundaries at query time
Cryptographic proof of what AI accessed
Instant revocation that actually works
Audit trail a High Court judge can understand
Public accountability without leaking internals

HAREUtil Workflow

Governed Project Query

Same question, different roles, different answers—all with evidence.

Ingest: Build Project Capsules

Project documents—EIAs, planning applications, grid studies, public submissions, internal memos—are ingested as capsules with policy bindings: classification, owner, jurisdiction, role permissions.

Capsule Builder

Query: User Asks AI Assistant

"Summarise all environmental mitigation commitments for marine mammals and note which are legally binding." Query arrives with user's role and jurisdiction.

PLAN mode

Filter: Arbiter Enforces Boundaries

Arbiter checks each capsule against user's role. MARA Case Officer sees internal risk memos. Journalist sees only public docs. Former consultant gets DENY on everything.

Arbiter

Redact: Sensitive Content Masked

Personal data in public submissions? Redacted. Legal privilege? Removed. Commercial terms? Hidden from public roles. AI only sees what role permits.

Partial Access

Answer: AI Generates Response

AI builds answer from permitted content only. Response includes source capsule IDs. Different role = different sources = different answer.

RAG + Llama

Log: Decision Artifacts Created

Every query produces evidence: actor, role, jurisdiction, docs accessed, docs denied, redactions applied. Cryptographic proof, not mutable logs.

Evidence Artifact

Role-Based Access

Same Question, Different Answers

The AI assistant sees the same project file differently depending on who's asking.

Role	Sees	Doesn't See
Developer PM	Everything except regulator-internal notes	MARA risk memos, ACP legal notes
Regulator Case Officer	Applications, EIA, submissions, internal risk notes	Developer legal privilege memos
Grid Planner	Grid studies, technical parameters, connection options	Internal legal/risk memos
NGO Lawyer	Public file, their own submissions	Internal memos, draft conditions
Local Resident	Non-technical summaries, visual impact, noise studies	Commercial terms, other residents' personal data
Journalist	Public docs, press releases, published decisions	Everything internal
Former Consultant	DENY on all queries	Access revoked—cryptographic proof of revocation

Use Cases

Where HAREUtil Applies

Offshore Wind & Renewables

Complex consenting across multiple agencies—MARA, planning authorities, grid operators, environmental regulators. AI assistance with legal boundaries baked in.

Grid Operations

Transmission upgrades, substation expansions, reliability planning. Role-based access from line supervisors to board members, all with audit trails.

Employee Self-Service

HR, payroll, benefits queries via AI—but employees only see their own records. Retirees get appropriate access. Former employees get instant revocation.

Public Accountability

FOIA-style requests answered by AI—but only from public records. Proof that internal documents weren't accessed. Transparency without leaks.

Regulatory Submissions

AI-assisted drafting of permit applications, environmental assessments, compliance reports—with proof of what source documents informed each section.

Judicial Review Defense

When decisions are challenged in court, produce cryptographic evidence that AI respected legal privilege, followed procedures, and accessed only appropriate documents.

Lighthouse Projects

Two Paths, One Platform

AtlanticCurrent (EU)

Offshore wind planning in Ireland—MARA, An Coimisiún Pleanála, EirGrid, environmental agencies.

Maritime Area Consent (MAC) applications
Environmental Impact Assessments
Grid connection studies
Public submissions and NGO engagement
Aligned with EU AI Act & MAP Act

US Public Power (Arizona)

HARE as informational spine for a public power utility—all departments, all documents, all employees.

HR, payroll, benefits self-service
Grid ops, safety, compliance
Customer service knowledge base
Retiree access with appropriate limits
Union-friendly trust boundaries

Why Now

The Regulatory Moment

Ireland's 20 GW Target

Ireland has ambitious offshore wind targets but projects face 4-8 year delays from consenting complexity. AI can help—but only if it respects legal boundaries.

Public Power Accountability

Public utilities serve ratepayers, not shareholders. AI assistants need to demonstrate trustworthiness to unions, regulators, and the public they serve.

EU AI Act Compliance

Energy infrastructure is high-stakes. AI systems in this sector need governance that satisfies EU AI Act requirements—with proof, not just policy.