IY2S510 AI Policy & Risk Matrix: 70% Assessment Guide (USW)
The IY2S510 Assessment 2 asks you to write an organisational policy, with guiding principles and a risk matrix, for introducing an AI system into a public service. It's worth 70% of your module mark (1,500 words). You're writing for Cooper's Chase Police Force and their PoliceAssist system. The assessment rewards coherent risk identification, well-grounded principles, and clear accountability—not an academic essay.
| Qualification | University of South Wales Level 5 |
|---|---|
| Module/Unit | IY2S510 Compliance and Risk Management |
| Assessment | Assessment 2: AI Policy & Risk Matrix |
| Weighting | 70% |
| Word Count | 1,500 words |
| Assessment Format | Policy document + risk matrix table |
| Submission Type | Individual coursework (written) |
What the assessment is really asking
Your task is to use information about Cooper's Chase Police Force and your own research into existing AI policies to produce a policy with a set of principles for implementing PoliceAssist. This is not an essay. You're answering three distinct questions:
- What principles should guide the use of AI in policing? Write 3–5 principles that are specific to this scenario and defensible with evidence.
- What are the operational and strategic risks? Identify 8–12 substantive risks and rate them on a matrix.
- How will accountability work? Show who owns what, and how the policy integrates into Cooper's Chase governance.
The marking rubric allocates marks as follows:
- Risk matrix – coherent, correct and covering all relevant risks: 35%
- Risk identification and mitigations: 30%
- Accountability and integration of the AI system: 15%
- Policy structure and professional presentation: 20% (implied)
Write it as a policy, not an essay
This is your biggest win. Students often submit a 1,500-word analytical essay with a table at the end. Examiners want a policy document. The structure should be:
- Executive summary (100 words): What is PoliceAssist? Why is this policy needed? What are the top 3 risks?
- Scope and objectives (150 words): Who uses it? What processes does it support? What are you trying to achieve?
- Guiding principles (300–400 words): 3–5 principles, each with a 50–80 word explanation. See below.
- Risk assessment and mitigation (400–500 words): Summary of your risk matrix, top 3–5 risks, and how they'll be managed.
- Accountability and governance (200–250 words): Who owns the system? How is it monitored? What's the escalation route?
- Risk matrix (table): 5×5 grid with 8–12 risks placed in cells.
- References (50–100 words): Cite your sources (Harvard style).
Total: ~1,500 words + matrix table (not counted in word limit). This structure shows the examiner you understand policy writing and risk governance, not just AI theory.
Build principles you can actually use
Your principles are worth defending. Each should be:
- Specific to policing and AI: Not generic corporate fluff.
- Actionable: Teams should be able to use it to make decisions.
- Supported by law or best practice: Reference the Data Protection Act 2018, Equality Act 2010, or policing ethics frameworks.
- Distinct from each other: Don't have overlap.
Example principles (write your own):
- Transparency by default. Officers and suspects must know PoliceAssist is being used. Explain how it works, what decisions it supports, and why (linked to GDPR Article 13 and Article 22 rights).
- Bias and fairness testing. Before and after deployment, the system must be audited for disparate impact on protected characteristics (age, ethnicity, disability). Linked to Equality Act 2010.
- Human oversight always. PoliceAssist recommends; humans decide. No automated suspect identification or arrest decisions.
- Data minimisation and retention. Collect only what's necessary, delete after defined periods. Linked to Data Protection Act 2018 principles.
- Accountability and redress. If someone is harmed by PoliceAssist, there's a clear complaints process and remedy.
Write 300–400 words total on principles. This signals you've moved beyond theory into practice.
Stuck on IY2S510 AI Policy & Risk Matrix: 70% Assessment Guide (USW)?
Send us your brief. A UK tutor will explain what the marking criteria are asking for and help you plan your answer. Your first consultation is free.
Find risks that belong to this scenario
A risk matrix is a heatmap that shows the severity score weighted by likelihood, usually on a scale of 1–5. Two rankings are required: one for severity of risks and one for likelihood. For PoliceAssist, relevant risks include:
| Risk Category | Example Risks | Why It Matters |
|---|---|---|
| Data & Privacy | Unauthorised access to suspect records; data breach; retention beyond legal limits | Police hold sensitive biometric and criminal history data. GDPR breach = reputational + financial + operational damage |
| Algorithmic Bias | System disproportionately flags suspects of certain ethnicity or socio-economic background | Violates Equality Act 2010; undermines public trust; could lead to unfair arrests or discriminatory policing |
| Accuracy & Reliability | Misidentification of suspects; false matches in facial recognition; system downtime | Wrong person accused = legal liability, harm to individual, operational failure |
| Human Override & Training | Officers ignore warnings because they distrust the system; insufficient training on when to rely on AI | AI becomes ineffective or is used incorrectly, leading to missed opportunities or false convictions |
| Accountability & Liability | Unclear who is responsible if PoliceAssist makes a wrong recommendation; vendor vs. police | Legal disputes; lack of accountability = lack of trust |
| Vendor Dependency | Vendor ceases support; licensing costs escalate; source code not disclosed | Operational continuity; cost control; ability to audit system |
| Public Perception & Legitimacy | Community concerns about surveillance; resistance from staff; negative media coverage | Erodes public trust in police; affects officer morale; could trigger FOI requests or regulatory scrutiny |
Your task: Identify 8–12 risks relevant to this scenario, rate each as Low/Medium/High for both Likelihood and Severity (or use 1–5 scale), and place them on your 5×5 matrix. Then—crucially—describe a mitigation for each risk (1–2 sentences).
Get the law right for policing data
This is essential. Accountability and integration of the AI system accounts for 15% of the mark. Your policy must reference:
- UK GDPR and Data Protection Act 2018: Lawful basis for processing personal data, data subject rights, consent, purpose limitation, data minimisation, storage limitation, and security.
- Equality Act 2010: Protected characteristics; duty not to discriminate; potential for indirect discrimination via algorithmic bias.
- Police Act 1996 & Police Reform and Social Responsibility Act 2011: Police governance and accountability; Chief Constable duties to the public.
- ICO guidance on AI and data protection: Available at ico.org.uk. Very current and credible.
- College of Policing guidance on ethics and AI: If available; shows you know the operational context.
Write 2–3 sentences per law or framework, linking it to your policy. E.g.:
"Under UK GDPR Article 22, automated decisions that have a legal or similarly significant effect on a data subject require safeguards. PoliceAssist supports human decision-making but does not make autonomous arrests, ensuring this requirement is met. Any facial recognition use must comply with Information Commissioner's Office guidance on automated processing and bias auditing."
Make the risk matrix coherent
Your 5×5 matrix is worth 35% of the mark. It must be:
- Properly scaled: Use the same scale for Likelihood (1–5) and Severity (1–5). Define what each number means (e.g., Likelihood 5 = >80% chance; Severity 5 = financial loss >£1m + criminal harm + regulatory action).
- Coherent: Risks with similar characteristics should sit in similar cells. Don't have all risks scattered randomly; that signals poor analysis.
- Complete: Cover all major risk categories (data, algorithmic, operational, reputational, vendor, legal).
- Colour-coded: Green cells (Low risk, 1–4), Amber cells (Medium risk, 6–9), Red cells (High risk, 12–25). This is standard ISO 31000 practice.
- Annotated: Each cell should contain a risk label (e.g., "Data Breach (L4, S5 = 20)"). Reference this in your narrative.
Example top risks and placements:
| Risk | Likelihood | Severity | Score | Colour | Mitigation (1–2 sentences) |
|---|---|---|---|---|---|
| Data breach / unauthorised access | 3 (Moderate) | 5 (Critical: GDPR fine, public trust, operational shutdown) | 15 (Red) | Red | Encryption at rest and in transit; role-based access control; annual security audit by independent third party; incident response plan in place |
| Algorithmic bias in suspect ID | 4 (Likely) | 5 (Critical: Equality Act breach, wrongful arrest, reputational) | 20 (Red) | Red | Bias testing against protected characteristics before and after deployment; external fairness audit quarterly; human review of all high-risk recommendations; training for all users on bias awareness |
| System misidentification (false matches) | 3 (Moderate) | 4 (Major: legal liability, wrongful conviction) | 12 (Red) | Red | Accuracy threshold set high; human verification of all matches; linked to legal checks before arrest; independent audit of error rates |
| Inadequate officer training | 3 (Moderate) | 4 (Major: misuse, liability) | 12 (Red) | Red | Mandatory training for all users; refresher every 12 months; competency sign-off; scenario-based e-learning modules; champion network in each police district |
| Vendor lock-in / cost escalation | 4 (Likely) | 3 (Moderate: budget pressure, operational constraint) | 12 (Red) | Red | Multi-year SLA with cost caps; escrow clause for source code; transition plan to alternative vendors; annual cost review against market |
| Public resistance / loss of trust | 4 (Likely) | 4 (Major: loss of community cooperation, legitimacy) | 16 (Red) | Red | Transparent communication with community groups; public FAQs on website; independent ethics review; independent audit findings published annually |
Present your matrix as a clean 5×5 table or graphic, then narrate the top 3–5 risks in your 400–500 word "Risk Assessment and Mitigation" section.
Accountability and integration
This is where many students lose marks. Write clearly about:
- Governance structure: Who approves the use of PoliceAssist? (e.g., Chief Information Security Officer, in consultation with Chief Constable and independent ethics board.)
- Roles and responsibilities: Who owns each risk? (e.g., Data Security Officer owns data breach risk; Operational Lead owns training risk.)
- Monitoring and reporting: How often is risk reviewed? (e.g., Monthly performance dashboard; quarterly risk review; annual external audit.)
- Escalation path: If a risk materialises, who is told and when? (e.g., Data breach → CISO → Chief Constable → ICO within 72 hours.)
- Integration into existing policy: How does this policy fit into Cooper's Chase's wider governance? (e.g., Links to Information Security Policy, Equality and Diversity Policy, Professional Standards Directorate.)
Write 200–250 words on this. It shows maturity and systems thinking.
Want a tutor to check your draft first?
Get clear written feedback on your own work before the deadline: structure, criteria coverage and referencing. Feedback on your own draft starts from £45, and tutoring from £32 an hour.
Mark-maximising checklist
Before submission, ensure:
- ☐ Format: Policy document (1,500 words) + risk matrix table (separate). Title page with name, ID, module code, date.
- ☐ Structure: Executive summary, Scope, Principles, Risk Assessment, Accountability, Matrix, References. Each section has a clear purpose.
- ☐ Principles: 3–5 principles, each specific to AI and policing, each defensible with law or best practice. No generic waffle.
- ☐ Risk matrix: 5×5 grid, 8–12 risks, each with Likelihood (1–5) and Severity (1–5), colour-coded, labelled clearly. Include a legend explaining your scale.
- ☐ Mitigations: Every risk has a 1–2 sentence mitigation that is specific and actionable.
- ☐ Law and policy: References to Data Protection Act 2018, UK GDPR, Equality Act 2010, policing governance frameworks. Citations link to your policy decisions.
- ☐ Accountability: Clear statement of who owns the system, how risk is monitored, and how it integrates into Cooper's Chase governance.
- ☐ Writing: Professional tone, no jargon without explanation, spellcheck, Harvard referencing throughout.
- ☐ Scenario-specific: You've read the Cooper's Chase briefing and evidenced understanding of their operational context (e.g., size, crime patterns, existing systems, staffing).
- ☐ Originality: Policy and matrix are your own work. AI was used only for research/brainstorm support, not text generation. No plagiarism.
- ☐ Word count: 1,500 ± 10% (1,350–1,650). Matrix table does not count toward limit.
Common mistakes to avoid
- Writing an essay instead of a policy. No: "The risks of AI in policing include data breaches, bias, and accountability failures. These are discussed extensively in literature…" Yes: "PoliceAssist will operate under the following principles: Transparency, Fairness, Accountability. Data breaches are managed by [specific mitigation]."
- Vague principles. No: "AI should be ethical." Yes: "Every use of PoliceAssist must be audited against the Equality Act 2010 and recorded in the police register, with annual published reports."
- Too many risks. 15+ risks suggests you haven't prioritised; 3 risks suggests you've missed major categories. Aim for 8–12, substantive and distinct.
- Risks without mitigations. A risk identified is not a risk managed. Every risk needs a 1–2 sentence action plan.
- Ignoring the law. This is policing and data. GDPR, Data Protection Act 2018, Equality Act 2010 are non-negotiable. Reference them explicitly.
- Unclear accountability. "The team will monitor it" is not enough. Name roles: "CISO (Jo Smith) owns data security; Ops Lead (Sarah Brown) owns training risk; both report to Chief Constable quarterly."
- Colour-coding without scale. If your matrix is red and green, explain what red and green mean (e.g., "Red = Risk Score 12–25, requires immediate mitigation").
- No scenario context. Generic policy = low marks. Show you've read the Cooper's Chase briefing (e.g., "As a 500-officer force with existing CCTV integration, Cooper's Chase can build on current digital infrastructure…").
- Plagiarism from online AI policy templates. You'll be caught. Use templates for structure only; write all content yourself.
Where students usually get stuck
Q: "How do I know if I've got the risk rating right?"
Use a simple scoring rubric. For each risk, ask:
- Likelihood: How probable is this in a police AI context, given the safeguards in place? 1 = Almost impossible; 5 = Very likely.
- Severity: If this risk materialises, how much harm? 1 = Inconvenience; 5 = Criminal harm, regulatory breach, loss of public trust.
- Score: Multiply (or sum) to get a priority score. 15–25 = Red (highest priority); 6–14 = Amber (monitor); 1–5 = Green (low).
Cite your reasoning in the narrative (e.g., "Algorithmic bias is rated Likelihood 4 because suspect databases are historically imbalanced, and Severity 5 because a bias case could trigger an Equality Act investigation, media backlash, and loss of community trust").
Q: "Should my policy recommend rolling out PoliceAssist or rejecting it?"
The brief doesn't require you to make a go/no-go decision. Instead, write the policy assuming deployment goes ahead, with mitigations in place. If your risk assessment shows the risks are too high to manage, say so explicitly (e.g., "If bias testing reveals a >5% disparate impact on ethnicity, deployment should be delayed pending redesign"). This shows critical thinking.
Q: "Can I use existing AI policy templates from Google, the Civil Service, or universities?"
Yes—as reference material and inspiration. But your policy must be written for Cooper's Chase Police Force and PoliceAssist. Copy-and-paste from templates = plagiarism. Use templates to understand structure, then write your own content.
Deadline coming up?
Book a 1-to-1 online session today. Tutors are available 7 days a week, 8am–11pm, and usually reply within 2 hours.
Frequently asked questions
How many risks should I include in the IY2S510 risk matrix?
The brief doesn't specify a number; however, 8–12 substantive risks is typical. Focus on risks directly relevant to introducing PoliceAssist into policing operations. Quality over quantity: each risk must have a clear impact, likelihood rating, and mitigation. Avoid padding; the marking rubric rewards coherence and understanding.
Which risk should I rate as highest priority in my matrix?
Data breach or privacy violation is often the highest for police AI. Algorithmic bias in suspect identification is also critical (high severity, high likelihood). Use your analysis of the Cooper's Chase scenario to justify your top 3–5 risks. Ensure each has both high impact and reasonably high probability.
Do I need to cite the Data Protection Act 2018?
Yes. Data Protection Act 2018 is UK law; UK GDPR is the core framework. For policing, cite these because they regulate how Cooper's Chase can use personal and biometric data. Law & policing risks are worth 15% (Accountability & Integration). Reference the ICO (Information Commissioner's Office) guidance on AI and data protection.
Can I use a 3×3 matrix instead of 5×5?
The brief doesn't prescribe matrix size. 5×5 is standard for ISO 31000 risk management and allows finer discrimination between medium and high risks. A 3×3 is simpler but risks clustering risks in the middle cell. Use 5×5 and label clearly (e.g., 1–5 for Likelihood and Severity). Justify your choice briefly in your methodology.
Who should own the overall risk register—the CEO, IT director, or compliance officer?
Make this explicit in your policy. For police AI, a Chief Information Security Officer (CISO) or Compliance Lead is typical. However, explain that accountability is shared: operational risk is owned by deployment teams, strategic risk by leadership. This shows systems thinking (a Distinction criterion).
Can I use AI tools to help me write this assessment?
USW provides guidance for students on the appropriate use of AI tools in assessment. The institution's position is that AI may support learning, planning and preparation, but your submitted policy and risk matrix must be written by you. You may use AI to brainstorm risks, check structure, or summarise source material. But the policy, principles, risk matrix, and all mitigations must be your own intellectual work. All risk identification, severity/likelihood judgements and strategic decisions must be yours.
How do I reference the Cooper's Chase information pack?
Treat the information pack as a client brief. Reference it once in your introduction (e.g., "Based on the Cooper's Chase Police Force briefing provided, this policy addresses the deployment of PoliceAssist, a decision-support tool for suspect identification and case prioritisation"). You don't need to cite it in-text repeatedly; instead, use it to inform your risk analysis and principles. Cite external sources (ICO guidance, law, academic literature) formally in Harvard style.
Key takeaway: This assessment rewards practical, law-aware policy writing over academic theory. Your examiner wants to see you think like a governance professional: identify real risks, ground decisions in law, and explain who is accountable. If you can do that in 1,500 words with a coherent matrix, you'll score well.
HND Assignment Help (hndassignmenthelp.com) is a UK-focused academic support service for students undertaking higher education qualifications, including Level 5 Computing and Business modules like IY2S510. If you'd like structured feedback on your draft policy, risk matrix, or accountability section, we offer one-to-one guidance to help you refine your work and maximise your mark. Visit HND Assignment Help for more support.
Get a free plan for this assessment
Tell us what you are stuck on and a HND Assignment Help tutor will send you a structure, the criteria to target and what to fix first – free, confidential, usually within the hour.
Prefer chat? Message us on WhatsApp. We never share your details.