AIRiskAware
What IsAI Concepts

What Is AI Ethics?

The principles and commitments that guide how AI is developed and used — fairness, transparency, accountability, human oversight, privacy, and safety. Distinct from compliance, but increasingly connected to it.

Ethics, governance, and compliance

AI ethics, AI governance, and AI compliance are related but distinct. Ethics asks: what should we value in AI — what principles guide our decisions about what AI to build and how to use it? Governance asks: what processes and accountability structures ensure those values are operationalised? Compliance asks: what specific legal and regulatory requirements must we meet?

An organisation can be technically compliant with data protection law while deploying AI in ways many would consider unethical. Conversely, aspirational ethics statements without operational governance mechanisms are ethics-washing. The most credible AI ethics programmes have all three: values articulated, governance to operationalise them, and compliance with relevant law.

Core AI ethics principles

Fairness
AI should produce outcomes that are equitable across groups and do not perpetuate or amplify discrimination based on protected characteristics.
Transparency
The operation of AI systems — how they make decisions, what data they use, what their limitations are — should be explainable to affected individuals and to oversight bodies.
Accountability
There should be a defined human or organisational entity responsible for the decisions AI systems make and the harms they may cause.
Human oversight
AI systems in consequential contexts should be subject to meaningful human review — particularly where outcomes significantly affect individuals' rights or interests.
Privacy
AI should use the minimum personal data necessary for its purpose, and individuals should retain meaningful control over how AI systems use data about them.
Safety and reliability
AI systems should behave reliably within their intended scope and have adequate safeguards against failure, misuse, or unintended consequences.

AI ethics in practice: what makes it real

A credible AI ethics programme has specific commitments, not generic values. This means: defining what fairness means for the organisation's specific AI use cases; setting explicit red lines (what the organisation will not do with AI even if legal); establishing an ethics review process for proposed AI deployments with genuine decision-making authority; and accountability mechanisms for violations.

The clearest signal that an AI ethics process is substantive: it has declined or required modification to proposed AI deployments. An ethics review that approves everything is not being applied rigorously. Regulators, investors, and civil society organisations increasingly ask for evidence of specific decisions made on ethics grounds.

Regulatory connections

AI ethics principles are increasingly embedded in regulation. The EU AI Act's prohibited AI provisions (subliminal manipulation, social scoring) operationalise ethics norms into hard law. The OECD AI Principles (2019) — adopted by over 46 countries — are increasingly referenced in national regulatory frameworks. Australia's AI Ethics Framework (2019) provided a voluntary baseline that has influenced the AI6 Essential Practices and government procurement standards.