AI Controversy Database
SCAND.Ai runs an autonomous 24/7 pipeline that detects, deduplicates, scores, and tracks AI-industry controversies across Hacker News, Google News, Reddit, RSS/ArXiv, Bluesky, and Polymarket. This page is the open distribution of that data.
Download
scand-ai-weekly.csv — weekly aggregates since 2026-03-15: topics detected and analyzed, category breakdown, peak/average noise, content volume, analyses completed. Refreshed continuously; every row carries its as_of_date and a methodology link.
| Coverage | 2026-03-15 → today (22,019 topics, 3,568 fully analyzed) |
|---|---|
| Granularity | weekly (Monday-start, UTC) |
| Format | CSV, UTF-8, header row |
| License | CC BY 4.0 — free to use with attribution to SCAND.Ai |
What the fields mean
- topics_created / topics_analyzed — new controversy topics detected that week, and how many carry a completed AI analysis.
- topics_safety … topics_other — category breakdown across the 8 controversy categories.
- peak_noise / avg_noise — the 0–100 noise score is a decay-adjusted measure of current loudness at observation time, not historical importance. Read the methodology before comparing across long periods.
- content_items_ingested / analyses_completed — pipeline volume (available from the start of operational metrics).
- as_of_date / methodology_url — when the row was generated and how to interpret it; always cite both.
Citing this dataset
SCAND.Ai AI Controversy Database, 2026-06-10. https://scand.ai/data (CC BY 4.0)
Per-topic detail (parties, timelines, forecasts, noise history) is available on each topic page and via the API. Quotable single statistics live on /stats.
See also: Methodology · Corrections policy · About