Moderation report
This report covers orders we received from authorities, content reported to Gloaming, content we acted on without a report, and how our automated tools performed. We publish one for each calendar year, within two months of the end of the year.
Orders from authorities
We received 0 orders to act against illegal content and 0 orders to provide information from the authorities of EU Member States in this period, under Articles 9 and 10 of the Digital Services Act. With no orders, there is no breakdown by type of illegal content or by Member State, and no median time to confirm receipt or to give effect to an order.
Notices received and acted on
| Category | Notices | Content removed | Account restricted | No action |
|---|---|---|---|---|
| Illegal content | 4,812 | 3,190 | 402 | 1,220 |
| Harassment or bullying | 9,204 | 5,878 | 1,133 | 2,193 |
| Hate speech or symbols | 3,551 | 2,704 | 611 | 236 |
| Nudity or sexual content | 12,660 | 10,942 | 308 | 1,410 |
| Violence or dangerous organisations | 2,087 | 1,540 | 295 | 252 |
| Spam or scam | 27,318 | 24,905 | 1,802 | 611 |
| Intellectual property | 1,944 | 1,201 | 58 | 685 |
| Self-harm or eating disorders | 1,376 | 902 | 77 | 397 |
| Total | 62,952 | 51,262 | 4,686 | 7,004 |
Every notice is counted once, in the category the reporter chose. "No action" means the content was reviewed and found not to break our Community Guidelines or the law.
- 1,208 of these notices came from trusted flaggers.
- 3,592 of the actions were taken because the content was illegal, and 52,356 because it broke our Terms and Community Guidelines.
- 24,117 notices were processed by automated means.
- Median time from notice to action: 11 hours. Notices alleging illegal content are queued ahead of the rest.
Content we acted on without a notice
| Type | Found by automated tools | Found by reviewers | Content removed | Reach limited | Account restricted |
|---|---|---|---|---|---|
| Spam or scam | 88,410 | 2,306 | 84,596 | 0 | 6,120 |
| Nudity or sexual content | 13,870 | 1,032 | 12,744 | 1,947 | 211 |
| Known illegal material | 1,117 | 0 | 1,117 | 0 | 0 |
| Violence or dangerous organisations | 2,960 | 418 | 2,204 | 1,090 | 84 |
| Self-harm or eating disorders | 1,422 | 655 | 731 | 1,346 | 0 |
| Total | 107,779 | 4,411 | 101,392 | 4,383 | 6,415 |
Each item is counted once, under the measure we took. 121 reviewers work across 14 languages. Every reviewer completes a four-week induction on our rules and the law before reviewing alone, takes a refresher each quarter, and has access to counselling.
Automated tools
We use automated tools for two purposes: detecting spam, and identifying content that matches known illegal material. 38% of removals in this period were made by automated tools; the rest were made by people.
To measure accuracy, reviewers re-checked a random sample of 5,000 automated removals. They confirmed 96.2%, an error rate of 3.8%.
Safeguards: automated decisions that restrict an account are reviewed by a person, the sample above is repeated every quarter and the tools are adjusted when the error rate rises.
Questions about this report go to our point of contact for authorities, listed on Contact.