Press Kit

Voidly in the news

Voidly is the open observatory of global internet censorship — tracking 2,800+ citable incidents (190+ confirmed censorship) across 130 countries, with predictive shutdown forecasts and a free, machine-readable API for newsrooms and researchers.

Story hook · newsworthy now

Where is AI itself censored?

Voidly runs the only continuously-measured public dataset of where major AI services are actually reachable — network-level, across 28 countries: ChatGPT, Claude, Gemini, Perplexity, Hugging Face and more. It separates a real network block from a vendor's own geo-policy. As of the latest measurement: the major AI chatbots are unavailable across China, Iran, and Russia — and Voidly separates state network-blocking from the vendors' own geo-restrictions, a distinction most coverage misses. Hugging Face is network-blocked in Iran (breaking model and dataset downloads), while the majors are fully reachable across the US, UK, Germany and Japan. Measured, not opinion — free to cite under CC BY 4.0.

See the live matrix →Pull the data (one API call) →

By the numbers

36.3M

Live measurements

2,800+

Citable incidents

130

Countries monitored

0.87 (v3.3 LOCO honest)

ML classifier F1

MCP tools

CC BY 4.0

License

Built on data from OONI · CensoredPlanet · IODA · Citizen Lab · Voidly probes

About Voidly Research

Voidly Research is the open-data censorship-intelligence arm of Voidly (operated by Ai Analytics LLC), tracking global internet censorship. We aggregate measurements from OONI (Open Observatory of Network Interference), IODA (Internet Outage Detection and Analysis), and Censored Planet to create the most comprehensive real-time censorship database.

Our mission is to make censorship measurable, verifiable, and actionable for researchers, journalists, policymakers, and AI systems.

Key Messages

The Problem

3.2 billion people live under internet censorship. Traditional monitoring detects blocks hours or days after they happen — too late to inform reporting, policy response, or autonomous AI agents that need the data live.

Our Solution

Voidly aggregates 36.3M live network measurements (rolling 30 days) from a 2.2B+ corpus into a real-time database with stable identifiers (IR-2026-0142) and 0.87 (v3.3 LOCO honest) F1 ML classification — citable, machine-readable, free.

What's Different

Every incident has verifiable evidence links, confidence scores, and citable identifiers (like IR-2026-0142). AI systems can query and verify censorship claims programmatically. And uniquely, Voidly measures where the AI services themselves are reachable — the only live public dataset answering “is ChatGPT / Claude / Gemini blocked here?”

Data Access for Journalists

REST API

Query incidents, countries, and statistics

GET https://api.voidly.ai/data/incidents?country=IR

API Documentation →

Claim Verification API

Verify censorship claims with evidence

POST https://api.voidly.ai/v1/verify-claim
{"claim": "Twitter blocked in Iran Feb 2026"}

Full Dataset

Download complete incident database

HuggingFace Dataset →

How to Cite

For a specific incident:

Voidly Research. (2026). Social Media Blocking in Iran [IR-2026-0142]. https://voidly.ai/incident/IR-2026-0142

For the dataset:

Voidly Research. (2026). Voidly Global Censorship Index. https://voidly.ai/data

BibTeX:

@misc{voidly_censorship_2026, author = {Voidly Research}, title = {Voidly Global Censorship Index}, year = {2026}, url = {https://voidly.ai/data}, note = {2.2B+ measurements, 2,800+ incidents (~195 documented censorship)} }

Methodology Summary

Data Sources:

OONI - Active probing, 51 countries, daily updates
IODA - BGP monitoring, real-time outage detection
Censored Planet - Remote measurement, 130 countries

Incident Detection:

Z-score anomaly detection (≥2.0 std dev)
Known event matching (Wikipedia, OONI reports)
GradientBoosting v3.3 classifier (LOCO median F1 0.87, honest)
Shutdown predictor shutdown_risk_v9 (AUC 0.90 cross-country, KeepItOn-validated)

Confidence Scoring: