Project Emet.
Investigative intelligence for journalists. Hebrew for truth — the activating word of the Golem. Built to surface what is hidden and verify what is claimed.
What it does
Give Emet a goal in plain language and it runs a multi-step investigation autonomously. An LLM reasons about what to do next while the agent searches entity databases, screens sanctions and PEP lists, traces corporate ownership chains, monitors global news, investigates blockchain transactions, and synthesises everything into a structured report — with PII scrubbed at publication boundaries and a full audit trail.
Built on the FollowTheMoney ecosystem
Emet sits on top of FollowTheMoney — OCCRP's data model for organised-crime investigations — and works directly with Aleph, OCCRP's investigative data platform. Federated search reaches across OpenSanctions, OpenCorporates, ICIJ Offshore Leaks, GLEIF, UK Companies House, SEC EDGAR, and GDELT. The name "Emet" (אמת) is the activating word inscribed on the Golem; the first letter, Aleph, is also the name of the platform Emet is designed to work with.
Two modes, one auditable boundary
In investigate mode all data flows freely so the AI can reason over complete information. Every operation is logged. In publish mode PII is scrubbed, sensitive data redacted, and every output sanitised before leaving the system. AI-generated findings are always flagged and require human verification. The system never fabricates evidence, impersonates sources, or publishes without human review.
Capabilities.
| capability | what it does |
|---|---|
| Federated search | 7 sources, parallel async fan-out, dedup + rate limiting + cache |
| Graph analytics | 7 investigative algorithms · NetworkX · export to Gephi/D3/Cytoscape |
| Forensic audit | gzip-compressed JSONL, SHA-256 verified — every tool call captured |
| Cross-session memory | file-based recall of prior findings, no database required |
| Interfaces | CLI · HTTP API · WebSocket streaming · MCP for Claude Desktop |
| Ethics | VALUES.json with five weighted journalism principles |
Roadmap.
This research is ongoing. If your work intersects with ours —