Open Research Platform

About Voidly

Structured censorship intelligence for researchers, journalists, and human rights organizations

Mission

An estimated 4.2 billion people live in countries where internet freedom is restricted (Freedom House, Freedom on the Net 2024). Traditional monitoring detects blocks hours or days after they happen — too late for journalists covering breaking events, human rights defenders documenting abuses, or people trying to access critical information.

Voidly makes censorship measurable, verifiable, and actionable — in near real-time.

We correlate data from multiple measurement networks into verified, ML-classified incidents with human-readable IDs, evidence chains, and confidence scores that researchers and journalists can cite.

Who This Serves

  • Journalists verifying censorship claims with citable evidence
  • Human rights defenders documenting internet shutdowns
  • Researchers studying internet freedom trends across countries
  • Developers building circumvention and internet-freedom tools
  • Civil society organizations monitoring government censorship
  • People in repressive information environments who need accurate, timely data

How We Know the Need

Iran — recurring total shutdowns

During the 2022 Mahsa Amini protests and subsequent unrest, Iran imposed multiple nationwide internet shutdowns. Our incident database documents hundreds of Iran-related censorship events, including WhatsApp, Instagram, and Signal blocking.

Myanmar — persistent platform blocking

Since the 2021 military coup, Myanmar has maintained blocks on Facebook, Twitter, and Wikipedia. Researchers tracking the situation need verified, timestamped evidence — not anecdotal reports.

Russia — expanding censorship infrastructure

Russia's TSPU (technical means of countering threats) has expanded from blocking individual domains to throttling entire protocols. Our probe network detects VPN blocking, protocol throttling, and DNS poisoning patterns that evolve weekly.

What's missing today

When journalists cover a breaking shutdown, they cite OONI raw measurements or anecdotal user reports. There's no single source providing classified, citable incidents with evidence chains and confidence scores — that's the gap Voidly fills.

How We Engage Users

Probe contributor feedback loop

Community probe operators report blocking patterns from their networks. A leaderboard and trust scoring system incentivizes sustained contribution. Contributors in high-censorship countries provide ground-truth signals that validate our ML classifier.

Researcher & journalist integration

Citable incident IDs (e.g. IR-2026-0142) were designed after feedback from researchers who needed stable references for academic citation. MCP server integration lets journalists query censorship data directly from their AI tools.

Open data as a feedback channel

All incident data is CC BY 4.0. When external researchers find discrepancies, they can file corrections through our API or GitHub. This creates a self-correcting loop independent of our team.

What's next

Structured user interviews with journalists and human rights defenders in target regions. Advisory relationships with civil society organizations. These are planned deliverables for the next funding cycle.

Why This Exists

Existing measurement projects are essential but leave a gap:

  • OONI generates raw probe measurements — doesn't classify incidents
  • CensoredPlanet runs remote DNS/HTTP tests — no real-time alerting
  • IODA tracks network outages — not targeted domain/platform blocking

Voidly bridges this gap: we correlate data from all three (plus our own 55 node probe network) into verified, ML-classified incidents with human-readable IDs (like IR-2026-0142), evidence chains, and confidence scores.

What We Provide

  • 55 globally distributed probe nodes generating proprietary measurement data
  • 317 verified censorship incidents with evidence chains
  • ML classifier (99.8% F1, internal validation) trained on 37K labeled samples
  • Shutdown forecast model (74.6% AUC, internal validation) — 7-day risk predictions across 126+ countries
  • Multi-source correlation: our probe network + OONI + CensoredPlanet + IODA
  • Real-time VPN accessibility testing across major providers
  • 10-year historical archive (1.6M records, 126+ countries)

Research & Open Data

Live Stats

Data as of March 2026

16.9M
Live Samples
99.8%
Classifier F1 (internal eval)
55
Global Nodes
126
Countries

Technical

AI Models
  • • Censorship classifier: 99.8% F1 (GradientBoosting, internal eval)
  • • Shutdown forecast: 74.6% AUC (XGBoost, internal eval)
  • • <50ms inference time
  • • Privacy-preserving: trained on aggregate data only
Infrastructure
  • 55 nodes, 6 continents
  • 99.8% uptime (self-healing)
  • • WireGuard + Stealth mode
  • • Zero manual intervention

Tools for Users in Censored Regions

Tools are free. Your anonymous usage data contributes to censorship measurement.

Transparency & Verification

Published Tools
Published tools (MCP server, probe app, API specs) are MIT licensed. Censorship data is CC BY 4.0.
github.com/voidly-ai →
Blockchain Experiments
Privacy verification experiments on Base L2. Contracts are live; full proof pipelines are under development.
View contracts →
Warrant Canary
0 warrants received. Updated monthly.
Verify status →

Operator & Capacity

Dillon Parkes

Founder & Lead Engineer

Built Voidly from the ground up — infrastructure, ML pipeline, probe network, and data ingestion. Also serves as CEO of Nexcom Media Group and as a board director of the Warriors Fund. Elected public official as Director of Montgomery County MUD No. 238 in Texas.

LinkedIn ↗
OperatorAi Analytics LLC (Texas)
What funding enables
1. Expanded probe network (55 → 80+ nodes in underserved regions)

Expanded measurement coverage in underserved regions (Sub-Saharan Africa, Central Asia, SE Asia). Beneficiaries: researchers and civil society. Metric: coverage in 80+ countries, ≤5min detection latency.

2. Faster incident classification pipeline

Sub-30-minute verified incident reports with evidence chains. Beneficiaries: journalists covering breaking censorship events. Metric: median detection-to-published < 30 minutes.

3. Community probe program expansion

Desktop probe app + leaderboard + measurement contribution. Beneficiaries: activists and researchers contributing local measurements. Target metric: 500+ contributors across 50+ countries within 12 months of launch.

Explore the data. Verify the methodology.

All censorship data is open. All tools are free.