AI Use Policy
We'd rather tell you what we actually do than make promises we can't keep. Here's the honest state of AI at Harbor Commons, as of late March 2026.
Current state
The Harbor Commons public site does not use artificial intelligence (AI) to generate content shown to visitors. All text on club pages — revenue figures, officer names, program descriptions — comes directly from IRS 990 XML filings with no AI-generated additions or summaries.
We use AI tools internally during development — for code review, drafting copy, and exploring patterns in the data. This is development tooling, not production content.
Plain-language definitions
Artificial intelligence (AI): software that helps analyze, summarize, classify, or generate content. Why it matters here: visitors should be able to tell which parts of Harbor Commons are direct public-record display and which parts are experimental or assisted tooling.
Large language model (LLM): a text-generating AI system such as ChatGPT or Claude. Why it matters here: Harbor Commons does not currently use an LLM to write organization-specific content shown on public pages.
Embedding search: a way to search by meaning rather than exact keyword match. Why it matters here: it can help people find similar program descriptions or narrative sections across filings, but it is still a search aid, not a substitute for reading the filing.
Embedding search (in development)
What we commit to
- \u2013We will not use AI to generate financial figures, officer names, or any factual claim about a specific organization.
- \u2013We will not train models on documents you submit to us privately.
- \u2013If AI-generated summaries ever appear on the public site, they will be labeled as such.
- \u2013We will update this policy before deploying any AI feature that touches visitor data.
Use of Harbor Commons by others
Harbor Commons is built for human readers and normal web indexing. Human-readable public pages are open to visitors and normal search indexing. Automated bulk extraction, mirroring, or use for model training or large-scale retrieval requires written permission.
Quoting, citing, and linking back to Harbor Commons pages is fine. Building silent mirrors, large-scale ingestion pipelines, or training corpora from the site without permission is not.
For machine-readable guidance, see llms.txt. In plain language: that file tells compliant AI and retrieval systems which parts of the site are meant to stay human-scale and which public pages are safer to reference.
It is guidance, not a lock. It helps make our intent legible to compliant systems, but it does not stop bad actors by itself.