Brazilian researchers scraped 2B+ Discord messages from 3,167 public servers using Discord's public API, covering 2015 to 2024, and publish the messages online
A Brazilian team used Discord's API to scrape 10% of its open servers. — Researchers published a massive database … Bluesky: @vortexegg.com , @cfiesler , and @josephcox Bluesky: @vortexegg.com : It's funny how all the extremism researchers in my circles are like “hello free data 👀” and all the critical tech scholars in my circles are lamenting how many very clear research ethics red lines this dataset violates [embedded post] Dr. Casey Fiesler / @cfiesler : Another week, another research ethics controversy. — TL;DR Researchers released a public dataset of 2B+ messages from 4M+ users on 3k+ “public” Discord servers. Usernames/IDs are anonymized. — But let's unpack this one... 🧵 — www.404media.co/researchers- ... Joseph Cox / @josephcox : New: Researchers published a massive database of more than 2 billion Discord messages that they say they scraped using Discord's public API. Released publicly so anyone can download www.404media.co/researchers- ...
Context & Ripple Effects
Discord’s exposure to large-scale collection was already visible in the Spy Pet scraping service, which claimed to archive data from thousands of servers for sale. Discord subsequently banned accounts linked to that operation, distinguishing enforcement against commercial harvesting from the broader accessibility of public-server data.
The newly published research corpus turns that access question into a research-ethics issue: anonymized account identifiers do not remove the sensitivity of a long-lived, searchable record of public conversations from millions of people.
First-order effects
- Discord users whose messages were posted in public servers now face a substantially more durable and easily reusable record of those conversations, despite the dataset’s reported anonymization.
- Researchers gain a large ready-made corpus for studying public-server activity, while Discord must confront how its public API can enable bulk collection at this scale.
Second-order effects
- The release sharpens the distinction between content being publicly viewable and participants reasonably expecting it to be mass-collected, retained, and redistributed—a boundary already tested by Discord’s response to data-selling scrapers.
- It may increase pressure on platforms and research teams to set clearer rules for API rate limits, archival access, dataset access controls, and disclosure practices rather than treating public availability as sufficient consent.
Third-order effects
- If public-platform research increasingly relies on bulk archives, platform governance will shift from moderating individual posts toward controlling the downstream lifecycle of accessible data: collection, retention, sharing, and reuse.
- The durable tension is that the same corpus can support investigation of harmful communities—an area where Discord has reported removing communities—while creating privacy and contextual-integrity risks for ordinary participants.
The trend: This is one data point in the widening public-data permission boundary, where technically open platform content is becoming contested once it is aggregated and redistributed at population scale.