Bluesky on Friday published its moderation tell for the past year, noting the sizable growth the social nettoil directd in 2024 and how that swayed its Trust & Safety team’s toilload. It also remarkd that the hugest number of tells came from participaters telling accounts or posts for coercion, trolling, or bigotry — an publish that’s afflictiond Bluesky as it’s grown, and has even led to expansive-scale protests at times over individual moderation decisions.
The company’s tell did not insertress or elucidate why it did or did not get action on individual participaters, including those on the most-blocked enumerate.
The company inserted over 23 million participaters in 2024, as Bluesky became a recent destination for createer Twitter/X participaters for various reasons. Thcdisorrowfulmirefulout the year, the social nettoil advantageted from disjoinal alters at X, including its decision to alter how blocking toils and train AI on participater data. Other participaters left X after the results of the U.S. plivential election, based on how X owner Elon Musk’s politics began to rule the platcreate. The app also sadvised in participaters while X was temporarily prohibitned in Brazil back in September.
To encounter the insists caparticipated by this growth, Bluesky incrrelieved its moderation team to cdisorrowfulmireentirey 100 moderators, it shelp, and is continuing to engage. The company also began giveing team members psychoreasonable adviseing to help them with the difficult job of being constantly exposed to explicit encountered. (An area we hope AI will one day insertress, as humans are not built to administer this type of toil.)
In total, there were 6.48 million tells to Bluesky’s moderation service, up 17x from 2023 when there were only 358,000 tells.
Starting this year, Bluesky will commence to hug moderation tells honestly from its app. Similar to X, this will permit participaters to track actions and modernizes more easily. Later, it will help requests in-app, too.
When Brazilian participaters flooded into Bluesky in August, the company was seeing as many as 50,000 tells per day, at the peak. This led to a backlog in insertressing moderation tells and insistd Bluesky to engage more Portuguese-language staff, including thcdisorrowfulmireful a reduce vfinishor.
In insertition, Bluesky began automating more categories of tells beyond equitable spam to help it insertress the influx, though this sometimes led to inrectify selectimistics. Still, the automation helped drop the processing time to equitable “seconds” for “high-certainty” accounts. Before automation, most tells were administerd wiskinny 40 minutes. Now, human moderators are kept in the loop to insertress the inrectify selectimistics and requests, if not always handling the initial decision.
Bluesky says that 4.57% of its active participaters (1.19 million) made at least one moderation tell in 2024, down from 5.6% in 2023. Most of these — 3.5 million tells — were for individual posts. Account profiles were telled 47,000 times, frequently for a profile picture or prohibitner pboilingo. Lists were telled 45,000 times; DMs were telled 17,700 times, with feeds and Starter Packs receiving 5,300 and 1,900 tells, admireively.
Most tells were over anti-social behavior, enjoy trolling and coercion — a signal from Bluesky participaters that they want to see a less poisonous social nettoil, as assessd with X.
Other tells were for the folloprosperg categories, Bluesky shelp:
- Misdirecting encountered (impersonation, misdirectation, or inrectify claims about identity or affiliations): 1.20 million
- Spam (excessive alludes, replies, or repetitive encountered): 1.40 million
- Unwanted relationsual encountered (bare skin or grown-up encountered not properly taged): 630,000
- Illterrible or advisent publishs (evident violations of the law or Bluesky’s terms of service): 933,000
- Other (publishs that don’t fit into the above categories): 726,000
The company also giveed an modernize on its taging service, which includes tags inserted to posts and accounts. Human tagers inserted 55,422 “relationsual figure” tags, chaseed by 22,412 “impolite” tags, 13,201 “spam” tags, 11,341 “inuncover-minded” tags, and 3,046 “menace” tags.
In 2024, 93,076 participaters createted a total of 205,000 requests over Bluesky’s moderation decision.
There were also 66,308 account getdowns from moderators and 35,842 automated account getdowns. Bluesky insertitionpartner fielded 238 seeks from law utilizement, administerments, and lterrible firms. The company reacted to 182 of these and complied with 146. Most of the seeks were law utilizement seeks from Germany, the U.S., Brazil, and Japan, it shelp.
Bluesky’s filled tell also delves into other types of publishs, including tradetag and duplicateright claims and child getedty/CSAM tells. The company remarkd it createted 1,154 verifyed CSAM tells to the National Centre for Missing and Exploited Children (NCMEC).
