Skip to content

Instantly share code, notes, and snippets.

@matthewdeanmartin
Created November 2, 2024 19:07
llm bot policy

Recommended policy compliance practices

Hachiderm's policy borderline "bot can only broadcast". So Riker Googling would probably pass, but an LLM bot like activist probably wouldn't.

Transparency

  • Mark account as bot in profile.
  • Optionally add #bot tag to each post, or other indication that it is AI generated.
  • If hosted on Hachiderm, add #hachybots to all posts.

Activity and server load

  • Post at most 1 post per hour. (Most common limit)
  • 5 scheduled posts per day (Hachiderm's policy)
  • Posts more often than 1 per hour should be unlisted.
  • Don't reply to a thread forever. (Hachiderm calls this a "doom spiral")

Degree of automation

  • No "uncurated news bots posting from third-party news sources"
  • Don't post without a human approving what the LLM has generated.

Interactions

  • Don't follow accounts marked with #nobot
  • "Bots cannot respond to hashtags, keywords, etc. without being tagged - e.g. bots that respond to user posts based on keywords and similar." Hachiderm's policy.

Moderation and compliance with other, non-bot specific rules

  • Bot should fetch server policy interaction accounts and hosting server and have a 2nd bot check if LLM content is compliant.
  • A bot should publish a profile page somewhere with its code of conduct, especially the verifiable parts (e.g. a report of posting frequency, etc.)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment