Deceit / Research Index
Epstein Files
Resource Index
Every database, search tool, archive, AI assistant, and journalistic platform built around the Jeffrey Epstein case files — verified July 2026. 24 resources across 7 tiers.
Reading rule. The DOJ warned this release may contain “fake or falsely submitted” material. A name in a file is not a finding. An allegation is not a conviction. Cross-reference across repositories. Images may be AI-altered. Verify before you amplify.
Tier 01
Flagship Databases
The two projects that became canonical. Both have their own Wikipedia articles. If you only use two tools, use these.
Jmail
Epstein’s files rendered as familiar apps.
Built in five hours by artist Riley Walz and Kino AI co-founder Luke Igel. Presents Epstein’s inboxes ([email protected], [email protected]) through Gmail, Google Drive, YouTube, Spotify, Amazon, and Wikipedia interfaces. Includes "Jemini," an AI search that rebuts DOJ claims that the files are unsearchable. 18.4M users; 1.4M files archived. Satellite projects: JPhotos, JFlights, Jamazon.
Open →EpsteinExposed
Network graphs, financial flows, and dossiers.
Built by a data engineer using the pseudonym "Eric Keller," a survivor of childhood sexual abuse. He quit his job to run it full-time. 2.15M documents indexed, 1,500 people catalogued. Network graph visualization, interactive flight log mapping, forensic financial analysis (Sankey diagrams of money flow), AI document summarization, person dossiers. Non-commercial, ad-free, community-funded. Excludes leaked material.
Open →Tier 02
Official Government Repositories
The primary sources. Every downstream tool indexes or derives from these. Start here if you want the raw, unmediated record.
DOJ Epstein Library
The canonical release site. 12 Data Sets.
The foundational corpus. Data Sets 1–8: FBI interview summaries and Palm Beach police reports (2005–2008). Data Set 9: emails and internal DOJ correspondence on the 2008 non-prosecution agreement. Data Set 10: 180,000 images and 2,000 videos. Data Set 11: financial ledgers and flight manifests. Data Set 12: late productions. January 30, 2026 release = 3M+ pages. The DOJ warned the release may contain "fake or falsely submitted" material.
Open →FBI Records: The Vault
Long-standing FBI FOIA vault.
The FBI’s Freedom of Information Act reading room for Jeffrey Epstein records. Predates the Transparency Act releases. Useful for cross-referencing against the DOJ Library.
Open →US Customs and Border Protection
CBP FOIA release on Epstein.
Customs and Border Protection FOIA records related to Jeffrey Epstein. A smaller, supplementary release.
Open →House Oversight Committee
33,295 pages (Sep 2) through photos/videos (Dec 18).
Multiple Dropbox and Google Drive drops between September and December 2025. Includes the 33,295-page September 2 release, 8,500 pages from the Epstein estate (Oct 17), 20,000 pages (Nov 12), and photo/video releases through December. SDSU LibGuides links to each folder directly.
Open →Epstein Files Transparency Act
H.R.4405. Signed November 19, 2025.
Passed the House November 18, 2025; Senate unanimous same day; signed by Trump the next day. The law that compelled the DOJ releases. Read it to understand what was required and what the DOJ admits it withheld.
Open →Tier 03
Curated Indexes
Directories of directories. Use these when you want a guided tour rather than raw search.
SDSU LibGuides
The most thorough curated directory.
Maintained by Katherine Holvoet, Scholarly Communications and Open Initiatives Librarian at San Diego State University. Links every official repo and public platform. Includes best-practices guidance on cross-referencing, OCR-error handling, and AI-image verification. The single best starting point for a researcher.
Open →Wikipedia: Epstein files
56-language article with a release chronology table.
Includes a chronology table of all releases and leaks (2024–2026) and a dedicated "Independent online databases" section. The "Conspiracy theories" section covers the "client list" narrative and the July 2025 DOJ memo that rebutted it.
Open →Tier 04
Community Search Platforms
Independent archives and search engines built by volunteers. Each has a different angle.
Epstein Archive
Auto-processed, OCR’d, searchable.
A public service project that automatically processes and OCRs publicly released documents for full-text search.
Open →Carstensen Epstein DB
Tracks which DOJ files were later deleted or modified.
Built by Tommy Carstensen, a Denmark-based data scientist and bioinformatician. Contains the original DOJ documents and identifies files that were later deleted or modified. Includes transcripts of handwritten survivor diaries and tracks real-world outcomes. Commits to redacting victim-identifying data. The closest thing to a version-tracked, diff-aware corpus.
Open →Epstein Network
Who appears in photos with whom.
Explores connections between people who appear in photographs from the DOJ-released Epstein Library dataset.
Open →Epstein OSINT Database
Verified reporting, court filings, survivor testimony.
A living database — part research tool, part public record — compiling verified reporting, court filings, survivor testimony, and public records connected to Epstein and his network.
Open →Jeffrey Epstein Library
Court docs, flight logs, client lists, declassified files.
A free public search engine dedicated to making officially released court documents, flight logs, client lists, and declassified files accessible.
Open →Tier 05
Open-Source Tooling
GitHub projects you can fork, audit, or self-host. The code is the documentation.
Epstein-doc-explorer
594 stars. AI-extracted relationship triples.
Analyzes the Epstein document corpus to extract structured information about actors, actions, locations, and relationships. 25,232 documents processed. BM25 keyword filtering, tag clustering, hop-distance analysis, community-editing schema. MIT license. Built with Claude Code.
Open →epstein-document-search
Page-level indexing with highlighted results.
A searchable database of Epstein court documents using Meilisearch. Page-level indexing for precise results, metadata extraction (case numbers, document IDs), filter by folder, highlighted search terms.
Open →nia-epstein-ai
Regex + vector + LLM with mandatory citation.
Open-source AI agent (live at epstein.trynia.ai). Hybrid architecture: traditional regex/grep for names, dates, identifiers + vector search for semantic queries + LLM orchestration that must cite sources. ~100M words indexed. 211 HN points. The HN discussion is a sharp critique of LLM grounding claims.
Open →Tier 06
AI / RAG Tools & Datasets
Chatbots and machine-learning datasets built on top of the files. Fast, but verify every answer against the source.
Nikity/Epstein-Files (HuggingFace)
4.11M rows, 341 GB, Apache Parquet.
Aggregated from the DOJ Epstein Library, House Oversight releases, and unsealed court docs. Includes extracted text from PDFs plus binary representations of audio, images, and videos. MIT license on processing. 4,013 downloads/month. The largest single ML-ready dataset.
Open →Document Detective
Google Gemini File Search over 20,000 files.
By MKWritesHere. Uses Google’s managed RAG (Gemini File Search API) to query 20,000+ Epstein files in ~4 seconds. Built on tensonaut’s HuggingFace dataset. No vector database setup required.
Open →ChatEpstein
Chatbot over ~300k documents.
A community-built RAG chatbot with access to roughly 300,000 Epstein documents for targeted search. Posted to r/Rag.
Open →Tier 07
Journalism & Leak Archives
Newsroom tools and leak archives. Some are proprietary to news orgs; others are public torrents.
Zeteo Searchable Docs
26,039 House Oversight docs, Google Pinpoint powered.
Published November 14, 2025 by Prem Thakker and Micah Lee. A searchable version of the House Oversight Committee documents using Google Pinpoint. Found dinners with Peter Thiel and Steve Bannon, chats with Ehud Barak, and extensive Trump references.
Open →Google Pinpoint
The underlying platform Zeteo used.
Free Google tool for journalists to search large document collections with AI. Search within troves of forms, handwritten documents, images, audio transcriptions, email archives, and PDFs. Transcribe audio and video. Transform tables to spreadsheets.
Open →BBC Verify
Systematically verifying released images.
BBC’s OSINT team is trawling through thousands of released images. Confirmed a Bill Clinton + Epstein concert photo from the 2003 Hong Kong Harbourfest Rolling Stones show by matching a lanyard and cross-referencing BBC reporting from the time.
Open →DDoSecrets Archive
439.88 GB consolidated. Torrents available.
Distributed Denial of Secrets bundled official releases (FBI, Interpol, DOJ, Bureau of Prisons, congressional, court) with the leaked Ehud Barak–Epstein email cache (obtained by hacking group Handala, 100,000+ emails) and files the DOJ released then withdrew. Torrents and magnets available. Searchable at libraryofleaks.org. Note: Handala has alleged ties to Iranian intelligence.
Open →