Prepared for the Safe Artificial Intelligence Fund

The civic ground-truth and verification layer for AI

GovData helps AI systems prove what a government actually decided, not merely generate a plausible summary.

Local-government decisions involving zoning, housing, infrastructure, energy, procurement, public safety, and regulation are fragmented across thousands of incompatible portals and millions of documents.

AI systems can retrieve and summarize individual records, but they cannot reliably determine what was proposed, amended, approved, rejected, postponed, or superseded unless the complete decision chain and chronology are preserved.

GovData normalizes these records into a source-linked government decision graph, creating responsible-AI infrastructure for citation-grade retrieval, temporal verification, interpretability, runtime safeguards, evaluation, compliance monitoring, and independent auditing.

13,000+ jurisdictions About 85% population coverage Over 20 years of history 22M+ source documents

Built by the original architect of Legistar.
Designed from the underlying legislative model outward, not from document summarization inward.

The Problem

Government information is public, but not AI-ready

Public records were designed for human access within individual jurisdictions, not for machine reasoning across thousands of governments and decades of changing decisions.

Fragmentation

No consistent source model

Meetings, staff reports, attachments, motions, votes, and final records are spread across disconnected vendor systems, archives, APIs, and custom portals.

Temporal Risk

The latest document may not be the final truth

A proposal can be amended, postponed, substituted, vetoed, reconsidered, or superseded. Correct interpretation requires chronology, not just semantic similarity.

Auditability

Summaries lose the evidence chain

When AI systems collapse multiple records into one answer, users often cannot see which motion, vote, attachment, or source record supports the conclusion.

The Solution

A source-linked decision graph for civic reasoning

GovData reconstructs the legislative process as linked data so that models can reason across the full history of a decision and return the primary evidence behind the answer.

Place → Government body → Meeting → Agenda item → Legislative matter → Attachment → Motion → Amendment → Vote → Outcome → Source

Identity

Resolves jurisdictions, governing bodies, officials, meetings, and matters across changing portals and vendor-specific identifiers.

Chronology

Connects repeated consideration, amendments, continuances, substitutions, votes, and later actions into a time-aware legislative history.

Provenance

Preserves links to agendas, minutes, staff reports, ordinances, exhibits, packets, and other authoritative public records.

Outcomes

Distinguishes what was discussed from what was adopted, rejected, postponed, amended, or superseded.

Foundation

Already operating at national scale

GovData is pre-revenue, but not pre-product. The collection, normalization, search, and update infrastructure is live across the United States and Canada.

13,000+ Jurisdictions cities, counties, school districts, and special districts
About 85% Coverage deduplicated population coverage across indexed government footprints
4M+ Events public meetings and related government events
22M+ Documents source records linked alongside 21M+ recorded votes

The dataset contains over 20 years of history, 189,000+ officials, and millions of relationships connecting meetings, matters, attachments, sponsors, motions, votes, and outcomes. It is refreshed continuously as public portals change and new records are published.

AI Safety

Why GovData belongs in the responsible-AI stack

The goal is not to make AI sound more authoritative. It is to make important civic answers more verifiable, reproducible, and resistant to plausible error.

Grounding

Reduce unsupported conclusions

Retrieve the complete decision history and require answers to cite the source records, motions, votes, and outcomes that support them.

Temporal Verification

Know what was true when

Determine whether a statement refers to a proposal, an intermediate action, a final adoption, or a decision that was later changed or superseded.

Independent Audit

Make answers inspectable

Allow regulators, businesses, journalists, researchers, governments, and the public to follow the evidence chain without trusting the model alone.

Not just better RAG

Generic retrieval can locate documents containing similar language. It does not inherently understand that an agenda item, amendment, substitute motion, roll-call vote, and later ordinance are different stages of the same decision.

GovData contributes the identity resolution, chronology, relationships, and outcome labels that allow retrieval to become evidence-based civic reasoning.

Designed for consequential use

Incorrect interpretations of local-government actions can affect development, compliance, investment, procurement, public reporting, legal analysis, and public trust.

The more consequential the use case, the more important it becomes to distinguish a fluent answer from a verified one.

Runtime Safeguards

Help AI know when to answer, qualify, or abstain

GovData does more than retrieve relevant records. It provides controls that help an AI system determine whether the evidence is current, consistent, complete, and sufficient for a reliable conclusion.

Supersession Detection

Find what replaced the earlier record

Identify when a proposal, ordinance, motion, approval, or policy was later amended, substituted, repealed, reconsidered, reversed, or superseded.

Conflict Detection

Surface inconsistent evidence

Flag conflicts among agendas, minutes, attachments, vote records, later actions, and other authoritative sources instead of silently selecting the most convenient document.

Evidence Completeness

Check whether the decision chain is complete

Determine whether the available record includes the motion, vote, outcome, chronology, and supporting source material needed to support a reliable answer.

Answer Validation

Test the conclusion against the evidence

Evaluate whether an AI-generated answer is actually supported by the cited records and whether the cited records support the specific claim being made.

Temporal Controls

Separate proposal from final action

Require answers to distinguish what was true at a particular point in time from what was ultimately adopted, rejected, delayed, or later changed.

Qualified Abstention

Do not manufacture certainty

Allow the system to state that the public record is incomplete, contradictory, or inconclusive rather than produce a confident but unsupported answer.

These safeguards turn GovData from a retrieval source into a runtime verification and accountability layer for AI systems operating in consequential civic, regulatory, legal, commercial, and public-sector settings.

Products

Three responsible-AI products from one maintained foundation

Each product uses the same source-linked decision graph, strengthening the underlying dataset while serving a distinct safety and governance need.

Product 1

GovData Verify

APIs and agent tools that return source-supported answers about government actions, including the relevant chronology, motions, votes, outcomes, and citations.

  • Citation-grade civic retrieval
  • Temporal and outcome verification
  • Structured evidence packages for agents
  • Conflict and supersession detection
  • Evidence completeness and qualified abstention

The core responsible-AI product.

Product 2

GovData Benchmarks

Evaluation datasets for measuring whether models and civic AI applications can interpret multi-stage government decisions accurately.

  • Hallucination and citation testing
  • Temporal reasoning benchmarks
  • Final-outcome classification
  • Provenance and evidence completeness

A measurable standard for civic-AI reliability.

Product 3

GovData Monitor

Continuous monitoring for organizations that need to know when local rules, approvals, obligations, or government positions change.

  • Regulatory and compliance alerts
  • Decision-change detection
  • Cross-jurisdictional monitoring
  • Primary-source audit trails

Governance infrastructure for ongoing operations.

Market

Safety infrastructure with immediate commercial value

Local-government decisions shape where projects can be built, how businesses may operate, what governments purchase, and which rules apply across thousands of markets.

AI Companies

Ground models in authoritative civic records

Model developers and AI applications need reliable government evidence, evaluation datasets, and audit trails for systems used in legal, regulatory, journalistic, commercial, and public-sector contexts.

Enterprise & Compliance

Know when local rules and approvals change

Multi-location operators, infrastructure developers, energy companies, real estate firms, and regulated businesses need verifiable cross-jurisdictional monitoring.

Government & Public Trust

Make civic AI inspectable

Governments, newsrooms, researchers, watchdogs, and residents need AI-generated explanations that can be traced back to the public record.

Local-government decisions influence trillions of dollars in public spending and private economic activity. GovData can become the trusted intelligence and verification layer for these decisions, first across North America and eventually across government systems worldwide.

Moat

The defensible asset is the maintained decision graph

Monitoring and summarization products validate demand. GovData is differentiated by the historical, cross-vendor, source-linked structure needed to verify outcomes and reconstruct decision history.

  • Thousands of public portals must be discovered and monitored
  • Vendor systems differ in APIs, schemas, archives, and attachment handling
  • Legislative identities must be resolved across meetings and changing source systems
  • Chronology and outcomes require domain modeling, not document matching alone
  • The pipeline must adapt continuously as portals, vendors, and records change

Why generic AI does not erase the moat

AI can accelerate extraction, classification, and summarization. It does not eliminate the need for a maintained national corpus, durable source identities, historical relationships, quality controls, and domain-specific outcome logic.

The models will change. The need for reliable evidence, provenance, and continuously maintained government data will remain.

Founder

Built from more than thirty years of domain experience

GovData is led by the person who originally modeled the legislative relationships the platform now reconstructs at national scale.

Darius Tajanko is a technology founder and municipal-data specialist with more than thirty years of enterprise software experience. He was the original architect of Legistar, beginning in the 1990s, and spent many years designing the legislative workflows and data structures used to manage government meetings, agendas, motions, votes, and public records.

He founded GovData to apply that experience to the fragmented world of local-government information. Darius designed and built the platform's ingestion systems, normalization framework, legislative decision graph, public portal, and AI-ready data infrastructure.

His advantage goes beyond collecting public documents. He understands how government decisions are introduced, amended, debated, voted upon, published, implemented, and preserved over time.

Founder-Market Fit

Generic scraping can collect pages. Generic AI can summarize text. GovData reconstructs the decision itself.

Darius Tajanko
Founder, hands-on software developer, serial entrepreneur, and original Legistar architect
Strategic Fit

Why GovData fits SAIF

GovData turns responsible-AI principles into deployable infrastructure for a consequential, fragmented, and globally important information domain.

Governance & Compliance Infrastructure

Operationalize responsible deployment

GovData provides the evidence, chronology, monitoring, and controls required when AI systems interpret government actions for regulated or high-impact use cases.

Trust Preservation

Make civic answers verifiable

Every conclusion can be traced through the decision chain to authoritative public records rather than accepted solely because a model sounds confident.

Alignment & Interpretability

Show why the AI reached its conclusion

GovData preserves the records, relationships, chronology, outcome logic, and conflicting evidence behind an answer so users can inspect how the conclusion was reached.

Decision-Making Tools

Give consequential users better evidence

Businesses, governments, regulators, researchers, journalists, and the public can make decisions using verified outcomes and primary-source audit trails.

Domain-specific model evaluation

GovData can test whether AI systems correctly reconstruct multi-stage government decisions, identify final outcomes, cite authoritative evidence, recognize superseded records, and abstain when the public record is incomplete or contradictory.

GovData can also serve as enabling infrastructure for civic-intelligence applications, including other companies in the broader responsible-AI ecosystem. The objective is not to replace every application, but to provide the normalized evidence and verification layer those applications need to produce more reliable results.

Investment

From national dataset to trusted AI infrastructure

GovData is seeking SAIF's early-stage investment and strategic guidance as part of the company's broader seed financing.

Product

Launch GovData Verify

  • Production verification API
  • Agent and MCP-compatible interfaces
  • Evidence packages and audit trails
  • Conflict, supersession, and completeness checks
  • Qualified abstention and outcome confidence scoring
Evaluation

Release civic-AI benchmarks

  • Hallucination and provenance tests
  • Multi-document temporal reasoning
  • Final-outcome verification
  • Public and private evaluation datasets
Market Validation

Recruit design partners

  • AI and agent developers
  • Compliance and regulatory platforms
  • Newsrooms and research organizations
  • Government and civic-technology teams

SAIF's value extends beyond capital. Geoff Ralston's experience helping deeply technical founders find product-market fit, sharpen their story, build high-value networks, and prepare for the next financing stage would be particularly valuable as GovData turns a difficult national data asset into a focused responsible-AI company.

The Close

AI should not guess what a government decided

GovData is building the evidence and safeguard layer that lets AI systems retrieve the right records, reconstruct the full decision, explain the evidence chain, cite authoritative sources, and abstain when the record does not support a reliable answer.

The Opportunity

GovData already has the national-scale data foundation, historical depth, legislative model, and operating pipeline.

The next step is to package that foundation into verification, evaluation, and monitoring infrastructure for responsible AI.

We are seeking an investment partner who recognizes that trustworthy AI depends on trustworthy underlying evidence.

Contact
Founder: Darius Tajanko
Company: GovData
Email: info@govdataconsulting.com
Phone: (312) 767-4004

Thank you for your consideration.
I built the legislative model that preserves the history of government decisions. Now I am building the infrastructure that allows AI to interpret those decisions accurately, transparently, and at global scale.