Responsible implementation mandate

Ethical AI Governance

AI opportunity is assessed alongside consequence. Our mandate is to improve organisations while protecting human accountability, workers, privacy, fairness, accessibility and security.

Human accountability

A named human owner remains accountable for consequential decisions. AI is not used to obscure responsibility.

Workforce impact before automation

Automation proposals consider affected roles, task redesign, accessibility, skills and transition needs. Productivity improvement is not treated as a licence for avoidable harm.

Privacy and data minimisation

Use the minimum necessary data, approved providers and defined retention. Sensitive information is not exposed to AI merely for convenience.

Fairness and accessibility

Higher-impact use cases are assessed for discriminatory effects, exclusion and accessibility barriers, with controls proportionate to consequence.

Bounded autonomy

Agents and automations receive only the permissions, tools and data they need. High-impact actions require explicit safeguards and appropriate human approval.

Evidence and challenge

Material outputs and claims are tested against authoritative evidence. Uncertainty, model limitations and assumptions are surfaced rather than hidden.

Security by design

Prompt injection, data exfiltration, tool misuse, excessive agency and unsafe outputs form part of AI security assessment.

Continuous governance

AI systems are inventoried, impact-screened, evaluated, approved with controls, monitored and re-evaluated when models, data, purpose or risk changes.

Applied to AI opportunity reviews

An opportunity is not recommended solely because it is technically possible or promises labour savings. Reviews should identify affected people and processes, data sensitivity, decision consequence, required human oversight, security exposure, accessibility impact, implementation dependencies and how benefits will be measured.