Legal
AI Processing Disclosure
Every report on CivicDeepDive is machine-generated. You deserve to know exactly how, and where it can be wrong.
What gets sent to the model
- The document you uploaded, or the public PDF at the link you pasted, plus any supporting file you attached.
- The agency name you entered, so the summary can be labeled correctly.
- Nothing else. Your email, payment details, and profile are never included in the prompt.
What the model produces
- A plain-English summary of the meeting.
- A list of resolutions with item numbers, dollar amounts, and votes where stated.
- A risk tier for each item, from clear through watch, warning, and critical.
- A transparency score based on how much was buried on the consent calendar or left without stated cost.
- Suggested follow-up steps you can save to your action planner.
Known limits
- Scanned or image-only PDFs read poorly and may produce thin results.
- Dollar figures that appear only in attached staff reports will not be captured.
- Risk tiers are heuristic judgments, not legal or financial conclusions.
- Names, vote tallies, and item numbers can be misread. Always verify before publishing a claim.
Human oversight
You remain the editor. Outreach letters, press kits, and complaints are drafted for you, but nothing is sent on your behalf without you reviewing and sending it yourself.
Training and retention
Your documents are sent to our AI provider only to produce your report. They are not used to train public models, and we do not sell or license them.
How to verify a finding
- Open the linked official record or the original PDF for the item in question.
- Match the resolution number and agenda item number shown in the report.
- Request the staff report from the clerk for any dollar figure that is not on the agenda text.
Questions about this page?
Send us a note through the contact form and we will respond by email.
Contact CivicDeepDive