AI Disclosure Policy

Last updated: August 5, 2026

Officer Responsibility Notice ReportMax AI generates draft narratives only. Every AI-generated sentence must be reviewed and verified by the reporting officer before any report is filed. AI does not replace officer judgment, supervision, or legal responsibility for the accuracy of official reports.

1. Purpose of This Policy

This AI Disclosure Policy explains how ReportMax uses artificial intelligence, what the AI does and does not do, how your data is protected during AI processing, and what obligations you have as the reporting officer. Transparency about AI involvement in law enforcement documentation is fundamental to our design.

2. What AI Does in ReportMax

ReportMax uses the Anthropic Claude API to perform a single, specific function: generating a first-person draft narrative based on the facts the officer types. The AI:

Reads the officer's tokenized input (see Section 4 for how PII is protected)
Generates a structured, first-person draft narrative organized in chronological order
Uses the officer's stated facts only — it does not consult external databases, body camera footage, dispatch records, or any other source
Inserts bracketed placeholders (e.g., [MISSING: time of incident]) when critical facts are absent rather than inventing them
Applies formatting appropriate to the selected report type (Incident Report, Traffic Stop, Arrest Narrative, Use of Force, DUI Supplement, Crash Report)

3. What AI Does Not Do

The AI in ReportMax:

Does not verify facts. The AI has no way to confirm whether the facts you enter are accurate. It generates a draft based solely on what you type.
Does not access any law enforcement database. The AI has no connection to NCIC, your agency's RMS, CAD systems, body camera footage, or any external data source.
Does not make legal conclusions. Unless explicitly provided by the officer, the AI does not determine probable cause, add charge elements, or draw legal conclusions.
Does not file or submit reports. No AI-generated content is transmitted to any agency system. The officer copies, reviews, edits, and submits the final report through their agency's own systems.
Does not learn from your reports. ReportMax does not use report content to train, fine-tune, or evaluate any AI model.
Does not guarantee accuracy. AI-generated text may contain errors, hallucinations, omissions, or awkward phrasing. Officer review is mandatory before any report is certified or filed.

4. Privacy Protection During AI Processing

Before any data reaches the AI, ReportMax's client-side tokenization automatically replaces personally identifiable information with opaque placeholder tokens:

Example: You type "I contacted John Smith, DOB 03/15/1990, near 123 Oak Street." Your browser automatically converts this to "I contacted [NAME_1], DOB [DOB_1], near [ADDR_1]." Only the tokenized version is sent to the AI. After the AI responds, your browser replaces the tokens back with real values before you see the draft.

This means the AI never processes or stores real subject names, addresses, dates of birth, phone numbers, or license plate numbers. The token-to-value mapping exists only in your browser's session memory and is discarded when you close or reset the report.

5. Mandatory Disclosure in Every Report

Every report generated by ReportMax includes a mandatory AI disclosure, both on-screen and in any printed or exported document. The disclosure reads:

"This narrative was generated by ReportMax AI based solely on the facts entered by the officer. AI-generated content may contain errors, omissions, or unclear wording. The officer must review every sentence, correct any errors, and certify accuracy before filing. AI does not replace officer judgment or legal responsibility."

This disclosure cannot be removed by officers or regular users. It is always present in the internal audit log.

6. Officer Certification Requirement

Before copying or printing any AI-generated narrative, you are required to certify the following statement:

"I certify that I reviewed this AI-assisted draft, compared it with my notes and investigation, corrected any errors, and accept responsibility for the factual accuracy and final content of this report."

This certification is recorded with your name and timestamp in ReportMax's audit log. It confirms that a human officer reviewed the AI draft and accepts responsibility for the final report. Certification does not transfer any legal responsibility to ReportMax or its operators.

7. Audit Trail

Every AI generation event is recorded in a tamper-evident, cryptographically hash-chained audit log. The log captures:

Officer identity and timestamp
Report type selected
Whether generation succeeded
Character count of generated output
Officer certification event (when the report is exported or copied)

Audit log entries cannot be altered or deleted by officers. This log provides an accountability record of all AI-assisted drafting activity.

8. Known Limitations of AI Drafts

Officers must be aware of the following before relying on any AI-generated draft:

Hallucination risk: AI models can generate plausible-sounding but factually incorrect statements. Every sentence must be verified against your notes and direct knowledge of the incident.
Entity detection gaps: Unusual name formats or abbreviations may not be detected by the tokenizer and could be passed to the AI as-is. Review the final output for any sensitive identifiers.
Context limits: Very long or complex incidents may result in condensed or incomplete narratives. Add missing details directly in the editable output before finalizing.
Tone and style: AI-generated prose reflects general law enforcement writing conventions and may not match your agency's specific style guide. Edit to conform to agency standards.
Legal sufficiency: AI output does not constitute legal review. Supervisors and legal counsel remain responsible for evaluating the sufficiency of reports.

Human-in-the-Loop

Every report requires active officer review and certification. AI assists drafting only — it never produces a final report.

Transparency Always

Every exported report permanently discloses AI involvement. This cannot be removed or suppressed.

Privacy by Architecture

Tokenization protects subject PII before transmission. The AI never sees real subject names or identifiers.

No AI Training

Your report content is never used to train, fine-tune, or evaluate any AI model.

9. Questions and Contact

Questions about how AI is used in ReportMax? Contact us at admin@reportmax.support. For agency-level AI governance documentation, contact us directly.