AI Playbook Assessment

UK Government AI Playbook compliance for responsible AI deployment - 10 principles and 6 ethical themes

UK Government AI Playbook Assessment

Assess UK Government AI Playbook compliance for responsible AI deployment, covering 10 core principles and 6 ethical themes.

Command

arckit ai-playbook <project or AI system>

Arguments

• system (required): AI system name or project identifier

Examples

arckit ai-playbook "Fraud Detection ML Service"
arckit ai-playbook "Benefits Eligibility Chatbot"

Purpose

Help UK government organizations assess compliance with the UK Government AI Playbook for responsible AI deployment.

Risk Level Determination

HIGH-RISK AI (requires strictest oversight)

Fully automated decisions affecting:

• Health and safety
• Fundamental rights
• Access to services
• Legal status
• Employment
• Financial circumstances

Examples: Benefit eligibility, immigration decisions, medical diagnosis, predictive policing

MEDIUM-RISK AI (significant impact with human oversight)

• Semi-automated decisions with human review
• Significant resource allocation

Examples: Case prioritization, fraud detection scoring, resource allocation

LOW-RISK AI (productivity/administrative)

• Recommendation systems with human control
• Administrative automation

Examples: Email categorization, meeting scheduling, document summarization

The 10 Core Principles

1. Understanding AI - Team understands AI limitations and realistic expectations
2. Lawful and Ethical Use - DPIA, EqIA, Human Rights assessment completed
3. Security - Cyber security assessment including AI-specific threats
4. Human Control - Meaningful human oversight and override capability
5. Lifecycle Management - Documented lifecycle plan from selection to decommissioning
6. Right Tool Selection - Problem clearly defined, alternatives considered
7. Collaboration - Cross-government collaboration and knowledge sharing
8. Commercial Partnership - Procurement includes AI-specific contract terms
9. Skills and Expertise - Team composition includes AI/ML, ethical AI, legal expertise
10. Organizational Alignment - AI Governance Board approval and SRO assigned

The 6 Ethical Themes

1. Safety, Security, and Robustness - Safety testing, fail-safe mechanisms
2. Transparency and Explainability - ATRS published, decision explanations available
3. Fairness, Bias, and Discrimination - Bias assessment, fairness metrics across protected characteristics
4. Accountability and Responsibility - Clear ownership, audit trail, incident response
5. Contestability and Redress - Right to contest AI decisions, appeal mechanism
6. Societal Wellbeing and Public Good - Positive societal impact, environmental consideration

Human Oversight Models

• Human-in-the-loop: Review EVERY decision (required for high-risk)
• Human-on-the-loop: Periodic/random review
• Human-in-command: Can override at any time
• Fully automated: AI acts autonomously (HIGH-RISK - justify!)

Output

Generates ARC-{PROJECT_ID}-AIPB-v{VERSION}.md with:

• Executive summary with overall score (X/160 points, Y%) and Go/No-Go decision
• 10 Principles assessment (each scored 0-10)
• 6 Ethical Themes assessment (each scored 0-10)
• Risk-based decision (HIGH/MEDIUM/LOW risk thresholds)
• Mandatory documentation checklist (ATRS, DPIA, EqIA, etc.)
• Action plan with priorities
• Links to existing ArcKit artifacts

Risk-Based Decision Criteria

• HIGH-RISK: MUST score ≥90%, ALL principles met, human-in-the-loop REQUIRED
• MEDIUM-RISK: SHOULD score ≥75%, critical principles met
• LOW-RISK: SHOULD score ≥60%, basic safeguards in place

Mandatory Documentation

ATRS (Algorithmic Transparency Recording Standard)

• MANDATORY for central government departments
• MANDATORY for arm’s length bodies
• Publish on department website
• Update when system changes significantly

DPIA (Data Protection Impact Assessment)

• MANDATORY for AI processing personal data
• Must be completed BEFORE deployment
• Must be reviewed and updated regularly

EqIA (Equality Impact Assessment)

• MANDATORY to assess impact on protected characteristics
• Must document how discrimination is prevented

Human Rights Assessment

• MANDATORY for decisions affecting rights
• Must consider ECHR (European Convention on Human Rights)
• Document how rights are protected

Prerequisites

MANDATORY (warn if missing):

• PRIN (Architecture Principles) - AI governance standards, compliance requirements
• REQ (Requirements) - AI/ML-related FR, NFR (security, fairness), data requirements

RECOMMENDED (read if available):

• DATA (Data Model) - Training data sources, data quality
• RISK (Risk Register) - AI safety risks, mitigation strategies

Guidance for High-Risk AI

STOP: Do NOT deploy without meeting ALL principles:

• Human-in-the-loop MANDATORY (review every decision)
• ATRS publication MANDATORY
• DPIA, EqIA, Human Rights assessments MANDATORY
• Quarterly audits REQUIRED
• AI Governance Board approval REQUIRED
• Senior leadership sign-off REQUIRED

Related Commands

• arckit atrs - Generate ATRS record (mandatory for AI systems)
• arckit dpia - Data Protection Impact Assessment
• arckit tcop - Technology Code of Practice (broader tech governance)
• arckit secure - Security assessment (AI-specific threats)

Resources

• AI Playbook
• ATRS Guidance
• Data Ethics Framework
• ICO AI Guidance