Original research · Version 1.0.0
AI News Observatory
A reproducible view of what our verified AI news archive actually contains. Every number below is recalculated from stable canonical stories and their recorded primary evidence.
Updated September 3, 2026 at 5:39 AM UTCData through 2026-09-03MethodologyDownload JSONDownload CSVCC BY 4.0
1037
Canonical stories
319
Last 7 days
1037
Last 30 days
355
Source domains
128
Updated stories
100%
Image coverage
Quotable findings
“AI Understanding's verified archive contains 1037 canonical AI news stories from 355 distinct source domains as of 2026-09-03.”
“1037 stories were published in the latest 30-day window in AI Understanding's archive.”
“128 canonical stories include a visible development history rather than a separate near-duplicate URL.”
“Research is the largest tracked topic in the archive with 457 matching stories; topic labels overlap.”
“100% of archived stories have a story-specific image recorded.”
Topic comparison
Rules-based topics overlap; do not add rows together.
| Topic | All stories | 7 days | 30 days | Source domains | Updated URLs |
|---|---|---|---|---|---|
| Research | 457 | 110 | 457 | 61 | 9 |
| Models | 195 | 61 | 195 | 65 | 27 |
| Safety | 148 | 49 | 148 | 78 | 26 |
| Policy | 122 | 52 | 122 | 102 | 28 |
| Chips | 89 | 19 | 89 | 61 | 20 |
| OpenAI | 59 | 26 | 59 | 30 | 19 |
| Anthropic | 41 | 18 | 41 | 32 | 15 |
| Google AI | 37 | 14 | 37 | 24 | 6 |
What this dataset can and cannot show
It describes AI Understanding’s archive—not the entire AI industry. Source-domain counts reflect recorded primary evidence URLs. Topic labels are reproducible rules, not a subjective ranking. Publication volume is not a measure of importance or market size. Review the methodology and record-level JSON before using a statistic.
From breaking news to lasting evidence
Follow immediate coverage on the news page and X, use the Observatory for stable evidence, then get the week’s most useful developments in one briefing.
Join the weekly newsletter →