We’re receiving about 3,000 reports/hour

Original Article ↗ Hacker News Discussion ↗

Bluesky’s rapid user growth is colliding with the hard realities of content moderation at scale, from handling CSAM and harassment to dealing with brigading, false reports, and politicized blocklists. Commenters debate technical and social approaches—LLM-based filters, web-of-trust systems, paid signups, TPM/KYC-style identity, and community-run labelers—while warning about privacy risks, echo chambers, and the limits of volunteer moderation. Many see moderation, not features, as the real determinant of whether Bluesky can stay usable and financially sustainable without repeating Twitter’s and Facebook’s failures.

Moderation Load and CSAM/Abuse Concerns

  • Users note an apparent surge of reports (including CSAM) as Bluesky’s user base explodes, and see this as the classic “UGC at scale” problem.
  • People discuss nuances of CSAM terminology and law: intent is to emphasize “abuse/exploitation,” but real‑world laws sometimes criminalize even “innocent” or “simulated” material.
  • Some stress that even teen‑to‑teen sharing can be abusive (e.g., bullying with leaked nudes), not just adult involvement.
  • Others mention how CSAM can be weaponized by hostile actors (e.g., planting it on free‑speech sites to trigger legal takedowns).
  • There are references to automated detection tools (PhotoDNA, Project Arachnid) and bad experiences on other platforms (e.g., Threads flooded with CSA‑adjacent content).

Bots, Spam, and Identity Friction

  • Many see bot/spam resistance as the decisive challenge; easy signup is believed to attract the same bot/propaganda operations seen on Twitter.
  • Suggested mitigations:
    • Limit reach of new accounts; invite/vouch systems; groupchat‑style smaller networks.
    • Paid signups (one‑time or recurring, fiat or crypto) to add friction; prior examples (Metafilter, WhatsApp, SomethingAwful) cited.
    • Proof‑of‑work / hashcash, hardware fingerprints (TPM), KYC/ID checks.
    • Reputation/karma systems that weight reports and throttle abusers.
  • Counterpoints:
    • Determined spammers and “bad actors” already pay for Twitter verification; $1 fees or KYC won’t stop serious operations and harm privacy/anonymous access.
    • Hardware IDs and KYC are seen as privacy‑toxic, linkable across services, and exclusionary to people without IDs or bank access.
    • Some believe the problem is fundamentally unsolved at scale.

Blocklists, Labelers, and Echo Chambers

  • Bluesky’s culture of “block and move on,” shared blocklists, and community labelers is praised as effective and user‑controlled.
  • Others warn that centralized or popular blocklists can be abused for personal crusades, leading to opaque blacklisting and “samethink” communities with little dissent.
  • Suggestions include:
    • Conditional or threshold‑based blocklists (e.g., auto‑block if multiple subscribed lists include the same account).
    • Making block reasons explicit and visible, to aid user judgment.
  • Skeptics argue this simply shifts power from corporate moderators to list maintainers, potentially re‑creating the same issues.

Centralization, Federation, and Web‑of‑Trust Ideas

  • Some argue Bluesky’s central moderation is inherently brittle and expensive compared to federated systems like Mastodon or more P2P models (e.g., Scuttlebutt, friend‑to‑friend networks).
  • AT Protocol supporters point to planned ecosystem of third‑party labelers, feeds, and community classifiers so users can compose their own moderation and ranking.
  • Alternative visions include:
    • Badge‑ or web‑of‑trust–based visibility, where groups grant membership and trust and feeds are filtered via those relationships.
    • Smaller, interest‑based communities and group chats as the “best” social network model.
  • Others counter that “unmoderated” spaces quickly become unusable due to spam and harassment; some level of moderation and curation is seen as unavoidable.

Volunteer vs Automated Moderation

  • Several commenters note that the wider internet has long run on unpaid volunteer moderators, but also that “moderation does not scale gracefully.”
  • There is skepticism that community labelers can match the reliability or resourcing of big‑tech moderation teams.
  • Some think LLMs are close to being able to enforce a platform’s “tone,” while others doubt automation can fully replace human review, especially for edge cases and serious content like snuff or CSAM.

Business Model and Sustainability

  • People question how Bluesky will fund growing trust‑and‑safety workloads while remaining ad‑free.
  • A linked company statement mentions a planned premium subscription (e.g., higher‑quality video, profile customization) but promises no algorithmic boost for paying users.
  • Some are cautiously optimistic this could work; others suspect that, like Twitter, influence or political/VC motives may subsidize the platform even if it’s not profitable.