# Twitter/X legal demands to remove content, 2012–2026: notes

Companion to `x-government-removal-requests.csv` (one row per half-year, 2012-H1 to 2026-H1). Compiled 2026-10-10.

## Sources

| Periods | Source | Type |
|---|---|---|
| 2012-H1 to 2021-H2 | Official Twitter data download for report #20 (Jul–Dec 2021), `Twitter_Transparency-Removal_Requests_Jul-Dec-2021.csv` and `..._Information_Requests_...csv`, in https://transparency.x.com/content/dam/transparency-twitter/download/2021-jul-dec/Twitter-Transparency-Report-20-Jul-Dec-2021.zip (linked from https://transparency.x.com/en/reports/transparency-reports). The file holds every country and period from 2012 onward. | Primary |
| 2019-H2 to 2021-H2 (headline compliance rate) | Period pages, e.g. https://transparency.x.com/en/reports/removal-requests/2021-jul-dec ("Big picture" section) | Primary |
| 2022-H1 | X blog, "An update on Twitter Transparency Reporting", 25 Apr 2023: https://blog.x.com/en_us/topics/company/2023/an-update-on-twitter-transparency-reporting | Primary (summary only) |
| 2024-H1 | X Global Transparency Report H1 2024, p.13: https://transparency.x.com/content/dam/transparency-twitter/2024/x-global-transparency-report-h1.pdf | Primary |
| 2024-H2 | X Global Transparency Report H2 2024, p.13: https://transparency.x.com/content/dam/transparency-twitter/2025/x-global-transparency-report_h2_2024.pdf (HTML version: https://transparency.x.com/en/reports/global-reports/2025-transparency-report) | Primary |

Every populated row comes from a primary source; none relies on secondary sources alone. Press coverage was used only to cross-check. The figures it confirms include 47,572 for 2021-H2, 72,703 / 46,648 / 9,364 for 2024-H1, the 99 US demands in 2018-H1 and 11,460 information requests at 40.3% for 2021-H2.

## Checks

- The CSV totals match the narrative on each period page for 2019-H2 through 2021-H2: 27,538; 42,220; 38,524; 43,387; 47,572 requests. Accounts specified also match, except for one difference listed below.
- The published headline compliance rates are 35, 31, 29, 54 and 51%. The request-weighted country rates in the CSV give 34.6, 31.0, 28.6, 54.0 and 51.5%, so the two agree.
- Older archive pages agree with the CSV wherever they give numbers. For example, 2014-H2 lists Turkey 477, Russia 91 and Germany 43; 2016-H1 lists Turkey 2,493, Russia 1,601 and 20,571 accounts.
- The CSV in the H2 2019, H1 2020, H2 2020 and H1 2021 data downloads gives the same per-period totals as the H2 2021 file, so the history was not revised between releases.

## Definitions and comparability

1. **Twitter "legal demands" (2012–2021) are not government-only.** They combine court orders with other formal demands "from both governmental entities and lawyers representing individuals". Twitter never published a government-only total. The CSV does split court orders from other legal demands.
2. **Compliance rate (Twitter)** is the % of demands where Twitter "withheld or otherwise removed some or all of the reported content". It includes removals for breaking Twitter's own rules (TOS), not only country withholding. A global rate appears only from 2019-H2 onward. Before that the official dataset leaves the field null, and older reports gave per-country rates only. Those cells are blank.
3. **Action rate (X, 2024)** is cases actioned divided by total requests (70.82% in H1 2024, 81.89% in H2 2024). The reports do not define "actioned" or say whether demands from private lawyers are counted. The header calls the data "government requests". Treat the 2021→2024 change in rate as indicative only, not like-for-like.
4. **2024 breakdowns use jurisdictions, not countries.** "EU" is a bloc and "All Others" is lumped together. The top-3 columns rank only the jurisdictions X lists separately, so a country hidden in "All Others" could outrank South Korea. X did not report US removal requests or accounts specified for 2024.
5. **Accounts specified** can be inflated by "batch" demands. Examples: Indonesia named 102,363 accounts in two demands in 2021-H1, and South Korea named 94,156 in 2021-H2.
6. **Information requests:** the 2012–2021 disclosure % is my own request-weighted average of the per-country rates in the official CSV, which are given to 2 decimals (±0.5 pp). It reproduces the published 40.3% for 2021-H2. The 2024 values are as published.

## Gaps (not interpolated)

- **2022-H1:** X published only a blog summary: "approximately 53,000" removal requests from governments, no global compliance rate, top countries named without counts, and "over 16,000" information requests.
- **2022-H2, 2023-H1, 2023-H2:** X published no report.
- **2025-H1, 2025-H2, 2026-H1:** as of 2026-10-10 the transparency.x.com home page still links to July–December 2024 as the latest report. I found no later global report.
- **Archived period pages for 2012–2019:** the interactive data tables on these pages no longer render, so only the narrative text remains. The official CSV covers the data.

## Discrepancies found

- 2020-H1 accounts specified: the page says 85,375, the CSV sums to 85,357. The CSV value is used.
- 2019-H1 Japan: the page says 5,127, the CSV says 5,144. The CSV value is used.
- 2015-H2 accounts: the page says 11,090, the CSV says 11,092.
- 2024-H1: the breakdown rows sum to 72,704, one more than the stated total of 72,703.
- 2024-H2: Social Media Today and Digital Information World described the 97,006 / 82% figures as "first half" 2024. The report itself covers July–December 2024.
- 2012-H1: there are only 6 requests and the top-3 includes ties (Greece 2, then several countries with 1).

## Separate source, not part of the series

Rest of World (Apr 2023, https://restofworld.org/2023/elon-musk-twitter-government-orders/) analysed Twitter's self-reported submissions to the **Lumen** database rather than transparency reports. From 27 Oct 2022 to about late April 2023 it counted 971 government and court requests: full compliance in 808 (83%) and partial compliance in 154, against roughly 50% full compliance before the takeover. Lumen covers only the requests Twitter chose to submit there, so these figures cannot be compared with the transparency-report series and are not included in the CSV.