How Hallkin uses AI
Version 2026-10-25-draft · [Hallkin legal entity name]
Draft under legal review. This text describes how Hallkin works today, but it isn't final. Before it is: Review by counsel. Voyage AI's training opt-out confirmed in writing; Anthropic zero data retention decided.
Hallkin uses AI to save volunteers time. It never decides anything; people do, and their decisions are recorded.
What the AI does
- Answers questions from your association's own documents only, quoting the passages it relies on and saying whether each point is a documented rule or an interpretation. If the documents don't answer, it says so and offers to ask the board.
- Checks ACC applications for missing items and against the rules, for the committee to read. It can be wrong and doesn't approve or deny anything.
- Sorts maintenance requests and rule concerns for the board, treating every concern as an unverified report.
- Drafts letters that a person edits and approves before anything is sent.
- Transcribes scanned documents, which a board member checks page by page before anyone can search them.
What it doesn't do
- Decide applications, find violations, set fines, or send anything without a person approving it.
- Give legal advice, or decide which of two conflicting rules controls.
- See other associations' documents, or information your role can't see.
Your information
- The AI providers (Anthropic for the language model, Voyage AI for search) receive only what each task needs, and may not use it to train their models.
- Questions you ask are deleted after 30 days; answers are kept only as a record of which sources were cited.
- Photos sent with rule concerns are never sent to AI.
If it gets something wrong
Check the cited passages, and ask your board. Every answer links to the document and page it relied on, so mistakes can be caught and reported.