Iran’s internet came back in May; three months on, nobody is measuring what came back
The blackout ran four months and eighteen days and produced the only hard numbers this desk has. The restrictions that outlasted it have produced none.
Iran’s nationwide blackout ran from 8 January to 26 May 2026 — four months and eighteen days — recorded as suppression of the 2025–26 protests and as a war measure. Restrictions were relaxed on 28 January; severe ones persisted. The most defensible cost figure is a concession rather than an estimate: Sattar Hashemi, Iran’s minister of communications, put the daily economic cost at $35.7 million, as reported by IranWire. A government minister costing his own government’s shutdown is worth more than a think-tank model, and is also the number most likely to be conservative. The wider figures in circulation — 53 consecutive days as of 21 April, described as the longest nationwide disruption ever recorded; connectivity at roughly 1 percent of pre-war levels; $70–80 million a day including indirect costs and about $1 billion cumulative by day 50; online sales down 80 percent — all reach readers through a single EJIL:Talk essay citing NetBlocks, IODA and NPR secondhand.
What is circulating instead is history. The Internet Society published shutdown-trend analysis in early August covering 2019 to 2025; the numbers are accurate and they are not about 2026. The same applies to CISA’s #StopRansomware advisory on Gunra (AA26-222A), roughly three weeks old, and to Amnesty’s Pegasus work — the hacking of a former MEP who had investigated Pegasus abuses, published 6 July, and the “Inside Pegasus” technical retrospective of 17 July — both of which resurface as though new. This desk found no in-window measurement of Iran’s residual filtering regime from NetBlocks, IODA, Filterwatch or Access Now, and no in-window statement from the Iranian ICT ministry.
This edition of the desk is running short by design. An eleven-query sweep across Iranian and Israeli operations, shutdowns, spyware, hacktivism and Gulf ransomware ended before it reached Arabic- and Persian-language outlets, Sahel connectivity, platform transparency reporting, AI-generated propaganda, or any vendor APT publication dated 28–31 August. Those are gaps, not findings.
Assessment: The pattern worth naming is citation laundering: a law blog cites a measurement outfit, a wire cites the law blog, and by the fourth reproduction a modelled estimate has hardened into a fact with no one left holding the methodology. Hashemi’s $35.7 million survives that test because it costs the speaker something to say. The 1-percent connectivity figure may well be right, but it should be read off IODA, not off a footnote. And the absence of any published measurement of post-May Iranian filtering is the real story here — a shutdown ends when the trackers stop counting, which is not the same as when users get their network back.