370+
Documented Entities
6,400+
Sourced Connections
95
Investigation Pages

01Source Hierarchy

Every claim on this site is sourced from publicly available government records. We use a strict source hierarchy — higher-tier sources override lower-tier when conflicts arise.

Tier 1 — Primary Government Sources

  • Auditor General of Canada — Independent audit findings tabled in Parliament. The gold standard of government accountability.
  • Health Canada Annual Reports — Official MAID statistics published annually. Direct government data.
  • Parliamentary Budget Officer (PBO) — Independent fiscal analysis of government spending and policy costs.
  • Hansard — Official verbatim record of House of Commons and Senate debates. What they said, on the record.
  • Supreme Court of Canada decisions — Binding legal interpretations from the highest court.
  • Canadian Human Rights Tribunal — Quasi-judicial findings on discrimination by the federal government.
  • Statistics Canada — Official national statistics agency data.

Tier 2 — Regulatory & Oversight Bodies

  • Commissioner of Lobbying — Registered lobbying contacts database (public).
  • Elections Canada — Campaign finance, donation records, riding results.
  • Ethics Commissioner — Conflict of interest findings and investigations.
  • Privacy Commissioner (OPC) — Privacy breach investigations and recommendations.
  • NSIRA / NSICOP — Intelligence and security review findings.
  • Provincial Auditors General — Provincial-level audit findings.

Tier 3 — Investigative Journalism & Court Records

  • CBC, Globe and Mail, Global News — Major investigative journalism (verified against primary sources).
  • Court records — Federal Court, provincial courts, tribunal decisions.
  • FOI/ATIP releases — Documents obtained through Access to Information requests.
  • Independent researchers — SecondStreet.org, Yellowhead Institute, Cardus, C.D. Howe Institute (verified against primary data).
02Cross-Referencing Standard

An entity or claim receives a "multi-source" flag when it appears in two or more independent datasets. This is a confidence indicator, not proof of wrongdoing.

  • Single-source claims are clearly labelled and attributed.
  • Multi-source entities (appearing in lobbying + donations + Hansard, for example) receive higher visibility in the network analysis.
  • Correlation ≠ causation — we state this explicitly. Lobbying is legal and registered in Canada. Appearing in multiple datasets means an entity is active in public life, not necessarily corrupt.
  • All subjects are entitled to due process. We document what the public record shows. We do not convict.

03How We Use AI

AI-Assisted Investigation

This site uses AI systems extensively for research, analysis, cross-referencing, and content generation. All data is sourced from public government records, parliamentary proceedings, lobbying registries, and official public records. AI assists in processing, correlating, and presenting this public information at scale. Every claim is backed by verifiable primary sources — the AI is the tool, the public record is the evidence.

  • No anonymous sources — every claim links to a named document, report, or public record.
  • No partisan framing — we hold all parties accountable (Liberal, Conservative, NDP, BQ). The data leads; we follow.
  • No external API dependencies — this site runs entirely on static files. No cloud services, no databases, no third-party AI.
  • No paywalls — all content is free and will remain free. Public records belong to the public.
  • No tracking — no analytics cookies, no user profiling, no data collection.
  • No cross-project speech — LIRIL narration on this site uses a Clara/female voice lock (js/liril-voice.js). Pre-recorded scene audio matches each page's scene-id. Other project voices do not bleed into accountability pages. Contract: site isolation contract.
  • Local model stability — website-only pipelines (web audit, statistical anomaly, normalize) run against this repo and write proof artifacts here. Other linked projects in the same runtime are separate; they do not share narration, doctrine, or page copy with this dossier.

04Model Research Watch

We review current public open-model families and maintain a research watchlist for TENET5. The immediate focus is on models that can strengthen reasoning, summarization, evidence grounding, and research monitoring without changing the site's evidence-first standard.

Active fit now

Current open-weight language models remain the strongest match for local TENET5 reasoning, summarization, and evidence review.

Under evaluation

Next-generation open models are being tracked for refinement quality and evidence-grounding workflows.

Research-only track

Frontier open research models are monitored for reasoning and evidence-grounding relevance, but are not used as general OSINT models.

05How to Verify Our Claims

Don't trust us — verify. Every source block on every page provides direct links to original documents. Here's how:

Government Sources

Legal & Oversight Sources

06Investigation Tools

Explore the data yourself using our open-source investigation tools — all running locally in your browser with no external dependencies.

Network Analysis

Interactive force-directed graph — 370+ entities, 6,400+ connections. Zoom, pan, filter by category.

OSINT Dashboard

Intelligence overview with scandal clusters, timeline, and data source listings.

Entity Profiler

Search and browse all documented entities with full profiles and connection maps.

Criminal Code Analysis

87 findings including 6 Rome Statute charges — mapped to specific Criminal Code sections.

Corruption Map

Multi-source entity overlap analysis showing where datasets converge.

Investigation Board

Visual pinboard connecting evidence threads across all investigations.

About This Site

Built by Canadian researchers. Powered by public records. AI-assisted research and analysis. No tracking. No paywalls. The data belongs to the people who paid for it — Canadian taxpayers.

Full Methodology & Transparency Statement → — data sources, OSINT pipeline, figure validation, LIRIL AI narration, open source codebase.