April 19, 2026
Safety filter or fun police?
Banned by Anthropic
Users say the safety filter went haywire—paying customers, novels, and tech talk got axed
TLDR: An independent site is logging 23 verified cases of Anthropic bans to push for fair, human reviews. Commenters are split between fury over false positives and no support, jokes about “kill” commands getting flagged, and cautions that platforms must protect users—while some predict local, open models will dodge this entirely.
A new independent site, bannedbyanthropic.com, is turning private AI lockouts into a public receipts party, and the comments are absolutely on fire. The project lists 23 GitHub‑verified reports from 15 countries, with users claiming Anthropic’s guardrails booted them for everything from Linux slang (“zombie,” “orphan,” “terminate”) to heartbreak metaphors to thriller‑novel plotting. One banned writer? A suspense scene allegedly set off the alarms. One admin? Apparently “kills” and “executions” (normal computer terms) tripped the wires. Drama status: activated.
Outrage leads the thread, with top voices fuming that even $200/month customers can’t reach a human for help. Others roll their eyes at the site itself—one commenter joked it “got banned from vibing” because it wouldn’t load on mobile. The memes flowed, especially the running gag that you can’t “talk Unix” without getting flagged. But there’s real debate too: some argue platforms must guard against self‑harm and abuse, even if it means false positives; skeptics say these case dumps can be noisy and misleading. The hottest take? Open‑source, run‑at‑home models could make corporate lockouts irrelevant soon. For now, the vibe is clear: people want a fair appeals button, not a brick wall—and this record is the community’s loudspeaker.
Key Points
- •“Banned by Anthropic” is an independent project collecting GitHub-verified reports from users who say they were banned or restricted by Anthropic.
- •The site outlines a four-step process: submit details, verify via GitHub OAuth, choose public/private visibility, and appear in the record or aggregate totals.
- •It states it never accesses users’ GitHub repositories or private data; verification is solely to confirm identity and reduce spam.
- •As of the snapshot, the site lists 23 total reports (20 public) from 15 countries, with role counts (2 founders, 5 developers, 2 researchers, 3 students, 9 users).
- •Public examples include bans reportedly triggered by Linux administration terms, metaphorical language about grief/burnout, a thriller novel plot point, and academic writing; some cases involve paid accounts.