AI for Internal Affairs
Assist the investigation.Never replace the investigator.
A practical legal, operational, and governance framework for using artificial intelligence to organise evidence, build timelines, prepare interviews, and strengthen administrative investigations while preserving confidentiality, procedural fairness, and independent human judgment.
AI should not determine credibility, infer deception, decide intent, select findings, or recommend discipline.
Assist, Do Not Decide
AI can organise and compare information. Investigators retain responsibility for judgment and findings.
Verify Every Output
No summary, quotation, timeline entry, or policy reference should be accepted without source review.
Protect the Case File
Use only approved systems with appropriate access, retention, and confidentiality controls.
Document the Process
Record where AI was used, what was reviewed, what was corrected, and who approved the final work.
Visual Investigation Workflow
Follow the investigation from complaint through appeal
Select any stage to see appropriate AI assistance, the investigator’s nondelegable responsibilities, and required controls.
Interactive IA Case Simulator
See what responsible AI assistance looks like
Select a complaint type and the available evidence. The simulator generates investigative support—not findings.
Build the training scenario
AI-assisted case workspace
Select the case information and run the simulator.
IA Prompt Library
Prompts designed to support—not prejudge—the investigation
Search, filter, and copy prompts for approved agency systems. Replace bracketed text and remove confidential information unless the platform is authorised for it.
Red Flags
Never delegate these decisions to AI
These functions create substantial reliability, fairness, labour, evidentiary, and governance risk.
Do not ask AI to decide who is truthful.
Do not infer motive or mental state from language or behaviour.
Do not allow AI to select sustained or not sustained.
Do not ask AI to recommend a penalty.
Do not upload protected case material to public tools.
Interactive Evidence Map
Investigative conclusions should be traceable to source evidence
Select an evidence node to see how it can corroborate, contradict, contextualise, or identify gaps in the case.
& ISSUES
AI Governance Scorecard
Measure whether the agency is ready for operational use
Check only controls the agency can presently demonstrate through written policy, assigned responsibility, training, contracts, or documented practice.
Foundational Case Law & Rights Library
Administrative investigations operate within an existing legal framework
This library identifies recurring doctrines for training and issue spotting. Agencies should confirm jurisdiction-specific law, labour agreements, statutes, and procedural protections.
Supervisor Review Tool
Check completeness without allowing AI to grade the investigation
This tool identifies whether expected components are present. It does not determine whether the evidence or findings are correct.
Downloadable Working Tools
Adaptable language for agency implementation
Open, edit, copy, and print the tools. Local legal and labour review remains necessary.
AI Use Review Checklist
Purpose, data, vendor, verification, audit, disclosure, and approval controls.
Investigator Verification Log
Document prompts, sources, inaccuracies, corrections, and final human determinations.
Prohibited Uses Provision
Sample policy language for credibility, findings, discipline, and confidential information.
Training · Policy · Implementation
Prepare the Internal Affairs unit before AI becomes routine.
Shield Public Safety Training helps agencies define approved uses, establish defensible controls, train investigators and supervisors, and preserve fair administrative processes.