For funders / Voidly Atlas
Fund better evidence.
Coverage, classification and public access are funding priorities, not promised outcomes.
Sources · dates · classification · confidence
Measurement coverage and censorship confirmation are different claims.
Voidly Atlas — Censorship Research
Making internet censorship measurable, verifiable, and actionable.
Voidly’s open internet-censorship research product · Voidly was founded by Dillon Parkes
The Problem
Internet censorship is expanding globally, yet remains difficult to measure systematically. Over 4 billion people live in countries where internet freedom is restricted, and the situation is worsening year over year.
Existing measurement networks — OONI, CensoredPlanet, IODA — produce raw technical data that requires specialized expertise to interpret. There is no single platform that aggregates these sources, applies machine learning classification, and makes the results accessible to journalists, researchers, policymakers, and civil society organizations in a structured, citable format.
What Voidly Atlas Does
- Operates 30+ probe nodes across 6 continents, testing a rotating subset of the Citizen Lab domain pool (live node, country and domain counts at /v1/probe/network)
- Aggregates that fleet with three external measurement networks into an index covering 130 countries
- Ingests from OONI (8 test types), CensoredPlanet (DNS + HTTP blocking), and IODA (outage alerts) on a 6-hour cycle
- ML classifier v3.3 auto-classifies incidents with confidence scores and evidence chains. Leave-one-country-out across 127 countries: mean F1 0.71, median 0.87 — the median is lifted by small-sample countries, and the censorship-heavy ones score materially lower. Per-country numbers and this caveat are published at /v1/classifier/info.
- 600+ citable censorship incidents — each with linked evidence and a human-readable ID for citation — within 7,500+ tracked incident records and 400,000+ evidence items. The remainder are connectivity-disruption signals and single-source detections held as suspected; neither is published as censorship.
- 10-year historical archive of 1.6 million records from the OONI corpus
- Free public API, MCP server (84 tools for AI integration), bulk exports (CSV, JSONL), RSS/Atom feeds, and webhook-based real-time alerts
- All data released under CC BY 4.0
What Exists Today
Voidly Network
30+ nodes, 6 continents
Incidents Database
600+ citable within 7,500+ records, 400,000+ evidence
Public API
REST endpoints, bulk export, webhooks
AI Integrations
MCP server, OpenAI action, HuggingFace datasets
Predictive Model
7-day shutdown risk forecasting
Country Coverage
130+ countries, 10 years of data
What Funding Enables
1. Expanded Voidly Network
Scale from 30 to 80+ nodes with priority deployment in underserved regions (Sub-Saharan Africa, Central Asia, Southeast Asia) where censorship measurement gaps are most critical.
2. Faster Incident Classification
Reduce time-to-verified-report to under 30 minutes. This requires higher-frequency ingestion, improved ML pipeline throughput, and automated evidence correlation across all three data sources.
3. Community Probe Program
Launch a volunteer-run Voidly Network targeting 500+ contributors in 50+ countries within 12 months. Desktop app (macOS, Windows, Linux) already built with auto-update infrastructure. Funding enables onboarding, documentation, and community management.
Team
Dillon Parkes — Founder & Lead Engineer
Also CEO of Nexcom Media Group and an elected public official (Montgomery County MUD No. 238, Texas).
Links
Website: voidly.ai
Report: voidly.ai/report
API Docs: voidly.ai/api-docs
GitHub: github.com/voidly-ai
Contact: hello@voidly.ai