Voidly
Research

Evidence for your reporting

Journalist toolkit

Everything a reporter on deadline needs from Voidly Atlas, in one shareable URL. Live numbers, embeds, RSS feeds, citation templates, methodology, contact.

All datasets, methodology pages, and API endpoints linked here are licensed CC BY 4.0 — free to quote and republish with attribution.

Citable numbers (updated every 15 min)

Citable censorship incidents
772
censorship or mixed; excludes suspected and draft
Citable with linked evidence
685
subset with at least one evidence permalink
Disruption records
5,705
network events, not labeled as censorship
Total incident records
10,081
includes suspected and disruption records
Evidence items
697,029
OONI + CensoredPlanet + IODA + Voidly probes
Last 24h / 7d
42 / 423
new incidents observed

Source: /data/incidents/stats (CC BY 4.0). This page revalidates every 15 minutes; unavailable fields show as —.

Embeddable widgets

Drop these iframes into a story or blog post. CSP-safe, responsive, no JS required on your end.

Live ticker (latest incidents)
<iframe src="https://voidly.ai/embed/atlas/ticker" width="100%" height="120" style="border:0" loading="lazy"></iframe>
Per-country sparkline (e.g. Iran)
<iframe src="https://voidly.ai/embed/atlas/country/ir" width="100%" height="220" style="border:0" loading="lazy"></iframe>
Replace 'ir' with any ISO 3166-1 alpha-2 country code.
Globe with active incidents
<iframe src="https://voidly.ai/embed/globe" width="100%" height="400" style="border:0" loading="lazy"></iframe>

RSS / Atom / JSON feeds

Citation templates

Citation templates

Every individual incident has its own citable URL with automatic BibTeX, RIS, and plain-text formats at/v1/atlas/cite/{incident-id}. For citing the dataset as a whole:

BibTeX
@misc{voidly-atlas-2026,
  title  = {Voidly Atlas: Real-time Internet Censorship Index},
  author = {Voidly},
  year   = {2026},
  url    = {https://voidly.ai/atlas},
  note   = {CC BY 4.0}
}
RIS
TY  - DATA
TI  - Voidly Atlas: Real-time Internet Censorship Index
AU  - Voidly
PY  - 2026
UR  - https://voidly.ai/atlas
ER  -
Plain text
Voidly. (2026). Voidly Atlas: Real-time Internet Censorship Index. Retrieved from https://voidly.ai/atlas (CC BY 4.0).

Model transparency

Every Voidly ML model exposes a public /info endpoint with training date, evaluation metrics, feature importance, and honest caveats. The model changelog page stitches them.

  • /atlas/transparency — single-page honesty dashboard: every /info endpoint inline with training date, LOCO metrics, top features, and caveats
  • /atlas/changelog — every deployed model + version + LOCO metrics
  • /atlas/models — model registry with transparency endpoints
  • /methodology — how we measure censorship + how the ML works
  • /sentinel/backtest — historical forecast vs observed (predicted-vs-real scatter)
  • /sentinel/calibration — live reliability diagram + ACI online calibration state
  • /atlas/findings — editorial deep-dives, including HONEST negative ML results (v3.4 regime-cluster, v3.5 TabPFN — both archived, neither promoted)
Latest model registry snapshot: 17 live models, generated 2026-06-21 04:06:51 UTC.

Where to dig deeper

Contact for interviews + corrections

Reporters: email press@voidly.ai for interviews, fact-checks, or data corrections. We aim to respond within 24h on weekdays.

Found a labeled incident that you can prove is a false positive? Use POST /v1/sentinel/report_miss or email us directly. We track these in our error queue and they feed the next retrain.