Characterizing Bluesky Content Moderation Service: From Automation of Service to Landscape of Harms

This research outlines the first large-scale empirical audit of the Bluesky Moderation Service (BMS) using its publicly available moderation logs from 2025. The study leverages the architectural transparency of decentralized platforms like Bluesky to overcome the typical opacity of content moderation systems on major social media platforms. It aims to investigate the BMS's operational mechanisms, its effectiveness in identifying harmful content, and the specific types of harms it targets.

Why it matters

This research provides critical insights into the operational characteristics and effectiveness of content moderation systems, particularly within the context of decentralized social platforms. Understanding these aspects is vital for developing robust governance frameworks and ensuring accountability in digital ecosystems, which can inform strategic decisions regarding platform design and regulatory compliance.

Key insights

• Empirical research on content moderation is typically hindered by the lack of transparency in major social media platforms.
• Decentralized platforms with public moderation logs, such as Bluesky, offer a new opportunity for independent audits of moderation systems.
• This study is the first large-scale audit of Bluesky's default moderation system, the Bluesky Moderation Service (BMS).
• The analysis utilized 10.6 million moderation labels from 2025 to evaluate the BMS.
• The research focuses on three core aspects: the degree of automation versus human oversight in BMS, its accuracy in detecting harms, and the range of harms it addresses.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.11373