Why there's no 'newly-censored this week' feed (yet) — an honesty note

The most-requested thing journalists want from a censorship observatory is a global 'what got newly blocked this week' feed. We built one, tested it, and took it down before it ever shipped — because the only version the current data supports would publish false accusations against democracies. The naive build (every country/domain pair whose first blocking measurement lands in the last 7 days) produced a top that read: 'Australia blocked sputnikglobe.com', 'United States blocked isiswomen.org' (a women's-rights NGO), 'Japan blocked isiswomen.org', 'UK blocked freespeechdebate.com'. None are real national blocks — they're single-probe anomalies (transient timeout, CDN redirect, one flaky vantage point), the noise any raw per-measurement signal carries. A feed telling the world the US just censored a women's NGO is worse than no feed. We couldn't just filter it: the confirmed national-block layer (>=3 independent networks) correctly excludes every one of those false positives, but it's regenerated as a periodic snapshot — every row carries the regeneration date, NOT the date the block began, so 'newly' has nothing to stand on; and the per-measurement evidence that does have timing mostly lacks an ASN, so we can't retro-fit a multi-network gate onto the timed rows. The timed layer is noisy, the clean layer is timeless, neither alone can answer 'what was newly, genuinely blocked this week.' What we ship instead: /data/incidents — curated, deduplicated, vetted events with evidence permalinks, plus its ?since= delta and RSS/Atom feeds. The real fix is a state-transition log: persisting a genuine first-seen timestamp on each confirmed national block so accessible->blocked transitions are recorded the day they're first confirmed across networks. That's on the roadmap. Until it exists, the global newly-censored endpoint stays unbuilt — on purpose. An observatory that names governments has to be right.