AI Usage Disclaimer

1. What AI powers MuniMind

MuniMind uses large language models (LLMs) and embedding models to answer natural-language questions, summarize documents, and surface relevant decisions and data. Current providers include Anthropic (Claude) for generation and Voyage AI for semantic search embeddings. Model versions change over time as we adopt newer and more capable models.

2. AI can be wrong

Even state-of-the-art AI systems produce errors. Known failure modes on the Service include:

• Hallucination — stating facts, sections, or citations that do not exist
• Stale data — answering based on records that have since been updated
• Misattribution — confusing decisions, applicants, or parcels with similar identifiers
• Out-of-scope answers — confidently answering a question the underlying data cannot actually answer
• Jurisdictional crossover — pulling a rule from one city when you asked about another
• Summarization loss — omitting a conditional, exception, or footnote that materially changes the answer

We run quality evaluations continuously and post measured pass rates internally, but no system is perfect. Treat every AI answer as a lead, not a conclusion.

3. What you should always do

• Follow the citations. Every substantive claim the AI makes should link to a source — open it, read the surrounding context, confirm the claim.
• Cross-check with the authoritative agency. For zoning rules, check the municipal zoning resolution directly. For decisions, check the issuing body's official record. For property data, check the assessor/recorder of deeds.
• Consult a professional. For legal decisions, use a licensed attorney. For surveys, use a licensed surveyor. For construction, use a licensed engineer or architect. For tax questions, use a CPA or tax attorney.
• Assume the data can be outdated. Our ingestion pipelines run on schedules; recent changes may not be reflected yet.

4. What the Service is NOT

• Not legal advice
• Not tax, financial, or investment advice
• Not a substitute for a licensed surveyor, engineer, architect, or appraiser
• Not a title search or title insurance product
• Not an official record of any government body
• Not suitable for emergency or time-critical safety decisions

5. How your chat data is handled

Your prompts and the AI responses are processed by our model provider(s) under commercial data-processing terms. Anthropic does not use API inputs or outputs to train their models. We retain chat logs for up to 90 days for quality debugging, unless you save a conversation or add it to a watch. See our Privacy Policy for full details.

6. Reporting an error

Found an AI answer that was wrong, incomplete, or misleading? Please tell us so we can improve the Service. You can use the in-app feedback button on any answer, or email team@munimind.com with the conversation ID (visible in the URL as ?conv=...).

7. Changes to this Disclaimer

We may update this Disclaimer as models, providers, and mitigations evolve. The "Last updated" date above reflects the most recent revision.

8. Contact

Questions about how we use AI: team@munimind.com