Bluesky has published its first full transparency report, offering a detailed look at how the platform handled safety, moderation, and regulatory issues over the past year.
The report brings together data on trust and safety actions, age assurance efforts, monitoring of influence operations, automated labeling, and other internal initiatives, marking the most comprehensive disclosure the company has released so far.
According to the report, Bluesky’s user base grew rapidly in 2025, increasing by nearly 60 percent. The platform expanded from 25.9 million users to 41.2 million by the end of the year. This figure includes both users hosted directly on Bluesky’s infrastructure and those running their own servers as part of the decentralized network built on the AT Protocol. During the same period, users created 1.41 billion posts, which accounted for more than half of all posts ever made on the platform. Media sharing also grew quickly, with 235 million posts containing images or videos.
Bluesky reported a sharp rise in legal requests in 2025, receiving 1,470 requests from law enforcement agencies, government bodies, and legal representatives. This marked a fivefold increase compared to the 238 requests the company received in 2024. While Bluesky had previously released moderation reports in earlier years, this is the first time it has combined moderation data with information about compliance, verification, and other operational areas in a single report.
User-generated moderation reports also increased over the year. Bluesky received 9.97 million reports in 2025, up from 6.48 million in 2024. The company noted that this rise closely matched its overall user growth during the same period. Around three percent of users submitted at least one report, with misleading content, including spam, making up the largest share. Harassment and sexual content followed as the next most common categories.
Within harassment reports, hate speech represented a significant portion, along with targeted harassment, trolling, and a smaller number of doxxing cases. Bluesky explained that many reports labeled as harassment involved antisocial behavior that was rude or disruptive but did not clearly fit into stricter categories like hate speech. Most sexual content reports were related to mislabeling, where adult material was not properly tagged, limiting users’ ability to control what they see. Smaller numbers of reports involved nonconsensual imagery, abusive content, and deepfakes.
Violence-related reports made up a relatively small share and included threats, glorification of violence, and extremist content. In addition to user reports, Bluesky’s automated systems flagged more than 2.5 million potential violations. The company highlighted a major reduction in daily reports of antisocial behavior after introducing a feature that reduces the visibility of toxic replies by placing them behind an extra click. Following this change, reports of such behavior dropped sharply, and overall reports per active user declined steadily throughout the year.
Beyond moderation, Bluesky said it removed 3,619 accounts linked to suspected influence operations, most of which were believed to be connected to Russia. The company also stepped up enforcement actions overall. In 2025, Bluesky took down 2.44 million items, including accounts and content, a significant increase compared to previous years. Temporary suspensions and permanent removals also rose, with many permanent bans targeting spam networks, impersonation, and attempts to evade bans.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
Despite the increase in removals, Bluesky emphasized that it prefers labeling content over removing users entirely. In 2025, the platform applied more than 16 million labels to posts, a sharp rise from the previous year. Most labels were used for adult or suggestive content, reflecting the company’s approach of giving users more control over what they see rather than relying solely on account bans.
