About

About AI Analytics

AI Analytics builds intelligence infrastructure for the open internet — measurement systems, OSINT pipelines, and cryptographic tooling used by journalists, researchers, and defenders across borders.

Three things you can verify right now

The method is published and externally corroborated. Methodology documents how a censorship measurement becomes a verified incident (cross-checked against OONI, CensoredPlanet, and IODA — three projects we do not run), how the federal ingest works, and the common method behind the accountability datasets. The rules live at /standards/.
The data is downloadable, source-labeled, and keyless. Every accountability dataset ships as static JSON (one manifest, with its license stated); the regulatory hub is available through a public REST API. Reuse terms vary by source. Check any number on this site against the data it links.
The code and datasets are public. github.com/voidly-ai (probes, MCP server, datasets) and HuggingFace (1.66M+ dataset downloads).

Independence, funding, ethics, and response SLAs: Governance.

What we ship

Voidly publishes a CC-BY-4.0 dataset of internet censorship measurements covering 200 countries. The Federal Regulatory Data Hub indexes 229datasets from government sources, regulators/SROs, and derived indexes on source-specific refresh cadences, with keyless access and source-specific reuse terms. The Swarm SDK ships post-quantum encrypted communications for autonomous systems to vetted defense partners. Nexcom operates 59 publications across 45 cities in Texas, Florida, and six other states. Also published: Verboten (a global banned-books index), eleven topic guides to the federal data, intelligence briefs, and technical writing on the infrastructure behind each project. Supporting these: OSINT ingestion pipelines, election anomaly detection, and digital-footprint reconnaissance tooling.

Internal tooling

Footprint Vault is in-house OSINT reconnaissance tooling: persistent cross-platform entity profiles from passive collection across 40+ sources, graph-based identity resolution, Certificate-Transparency and BGP/ASN monitoring, and stylometric fingerprinting, feeding entity attribution back into Voidly. It is not publicly available — no hosted app, account, or API (methods write-up). Alongside it: a real-time OSINT ingestion pipeline (2.4M posts/hour) and election anomaly detection across voter, turnout, and campaign-finance data.

How we work

Self-directed. We pick our own targets, and we don't take work that conflicts with the open-data mission.
Verifiable claims. Numbers on this site link back to their source — Voidly's live dashboard, GitHub repos, or the linked writing post — so a reader can check.
Anonymous by design. Operator anonymity is a deliberate security posture for people who document censorship and surveillance infrastructure, not a hedge — the published method, open data, and public code are the credibility signal, and each can be verified without knowing a name.

Stack

OSINT pipeline: Kafka · TimescaleDB · Postgres
ML / NLP: XGBoost · ONNX Runtime · spaCy
Cryptography: ML-KEM-768 · X25519 · Double Ratchet
Voidly probes: 37+ nodes · 200 countries
Regulatory API: Cloudflare D1 · Workers · Next.js
Frontend: Next.js 14 · TypeScript · Cloudflare Pages

Where we operate

Probes run from 37+ vantage points spanning every continent. Our data is consumed by reporters and researchers in the EU, UK, US, Latin America, MENA, sub-Saharan Africa, South Asia, and East Asia. We document English-language sources by default and accept reports in any language we can machine-translate or have a partner verify.

Probe coverage: 200 countries · 6 continents
Working hours: Async across timezones
Response SLA: 24–48 hours
Reuse: Stated per dataset / source

Open work

api.ai-analytics.org — keyless Federal Regulatory Data Hub API (229 datasets; reuse terms vary by source)
github.com/voidly-ai — Voidly's 22-repo public org (probes, MCP server, datasets)
github.com/AI-Analytics-org — organization profile
huggingface.co/emperor-mew — datasets (1.66M+ downloads)
ai-analytics.org/writing — 526 long-form technical articles on censorship measurement, federal regulatory data pipelines, and post-quantum cryptography

How to cite us

Site: AI Analytics. https://ai-analytics.org (accessed YYYY-MM-DD).

A dataset: AI Analytics. (2026). [Dataset name] [Dataset]. https://ai-analytics.org/voidly/[slug]/ ([license]).

Per-dataset citation blocks appear on the dataset pages; the press kit carries the quotable fact sheet with as-of dates.

Federal Regulatory Data Hub: cite the catalog and the underlying source dataset used. Reuse terms vary by source. US federal employee-authored works are generally public domain under 17 U.S.C. § 105, while non-federal and SRO source material retains its source terms. AI Analytics-authored metadata and normalization are CC0 only where explicitly stated and applicable.