# Curriculum — Using AI in Law Enforcement ## 1. AI, Policing and Investigation: What Has Changed? * Generative AI, machine learning and AI-enabled investigative tools * How AI is already entering policing and investigative practice * Where AI can enhance professional capability * The difference between **assistance, augmentation and automated decision-making** * Why apparently convincing AI outputs can be wrong **Case study:** A recent example of AI use that created investigative, legal or evidential concerns. --- ## 2. Opportunity: Where Can AI Improve an Investigation? Practical exploration of legitimate applications, including: * Reviewing and organising large volumes of information * Research and information discovery * Summarising documents and material * Identifying themes, patterns, connections and inconsistencies * Supporting analytical thinking and hypothesis development * Translation and transcription * Preparing briefings, chronologies and working documents * Supporting administrative and repetitive investigative activity * Using AI to release investigator time for higher-value professional judgement **Exercise:** Delegates identify where AI could add value across an investigative scenario — and where it should not be introduced. --- ## 3. The Critical Boundary: Assistance or Evidence? Explore one of the central questions for investigators: **Is AI helping me understand the evidence — or starting to construct it?** * Original material versus AI-generated material * Evidential provenance and continuity * AI-generated summaries and interpretations * Creating or filling gaps that do not exist in the original evidence * Alteration and enhancement of images, audio and video * Synthetic material and deepfakes * Distinguishing investigative leads from evidence * Maintaining an auditable route back to the original source **Immersive challenge:** Delegates receive an AI-assisted investigative product and must determine what can safely be relied upon, what requires verification and what may have compromised evidential integrity. --- ## 4. When AI Gets It Wrong * Hallucination and fabrication * False citations and invented sources * Misinterpretation of context * Automation bias * Confirmation bias * Embedded bias within data and models * Overconfidence in apparently authoritative outputs * Why human checking alone is not always an adequate safeguard **Case studies:** Examine examples where inappropriate AI use resulted in inaccurate, misleading or potentially damaging professional outputs. --- ## 5. Law, Ethics, Disclosure and Accountability Led with input from legal practitioners. * Who remains accountable when AI contributes to a decision? * Fairness and procedural integrity * Disclosure implications * Recording AI use * Transparency and explainability * Data protection and privacy * Confidential, sensitive and legally privileged information * Intellectual property and third-party systems * What might a prosecutor, defence practitioner, court, regulator or inquiry subsequently ask? **Challenge question:** > *Could you explain and defend your use of AI six months later, under professional or legal scrutiny?* --- ## 6. Information Security: What Are You Giving the AI? * Public versus organisational AI systems * Sensitive operational information * Personal and special-category data * Intelligence and investigative material * Witness, victim and suspect information * Commercial AI platforms and data retention * Organisational policies and approved systems * Prompt design without unnecessary disclosure of sensitive information **Exercise:** Delegates decide what information can and cannot safely be entered into different AI environments. --- ## 7. Verification: Trust but Never Assume A practical approach to AI-assisted work: **Source → AI Output → Verify → Corroborate → Professional Judgement → Record** Participants practise: * Checking output against primary material * Identifying unsupported assertions * Checking quotations, citations and references * Separating fact from AI inference * Recognising uncertainty * Testing alternative hypotheses * Recording verification The principle is: **AI output is something to examine — not something to believe.** --- ## 8. Making the Decision: Should I Use AI? Introduce a practical **AI Investigative Decision Framework**. Before using AI, ask: ### PURPOSE What am I trying to achieve? ### AUTHORITY Am I permitted to use AI for this purpose and information? ### INFORMATION What am I putting into the system? ### EVIDENCE Could its use create, alter, contaminate or obscure evidence? ### RELIABILITY How will I verify the output? ### BIAS Could AI or my use of it reinforce assumptions? ### TRANSPARENCY Can I explain what the AI did and what I did? ### RECORD What needs to be recorded? ### ACCOUNTABILITY Who makes the final professional decision? --- # Final Immersive Exercise — The AI-Assisted Investigation Delegates work through a developing investigation in which AI presents both genuine opportunities and significant risks. At successive stages they must decide: **USE IT → USE WITH CONTROLS → VERIFY FIRST → ESCALATE/SEEK ADVICE → DO NOT USE** New information and consequences are introduced in response to their decisions. The exercise requires participants to balance: **Speed | Efficiency | Investigative Opportunity | Evidential Integrity | Security | Fairness | Legality | Professional Judgement** There may deliberately be situations where using AI is the best professional decision — and others where **not using it is the best decision**. # Assessment Participants complete an assessed scenario requiring them to demonstrate that they can: * Identify appropriate opportunities for AI within an investigation. * Recognise legal, ethical, security and evidential risks. * Distinguish AI-assisted investigative activity from the creation or alteration of evidence. * Verify and challenge AI-generated outputs. * Apply appropriate safeguards and escalation. * Record and explain how AI has been used. * Make and justify a defensible professional decision. Successful completion therefore demonstrates more than knowledge of AI. It demonstrates the ability to exercise **professional investigative judgement about AI**. ## Core Principle **Use AI to enhance the investigator — not to replace professional judgement or construct the evidence.**