AI Ethics, Compliance, and Regulatory Landscape
Principles, law, and the operating bridge
Candidates must connect ethical principles to legal obligations and to operational controls. Know the major regulatory patterns rather than memorizing statute numbers: risk-based, tiered obligations (prohibited practices, high-risk systems with conformity and documentation duties, transparency duties for certain AI interactions, and general-purpose model obligations); sector rules layered on top (financial services, health, employment, education); and cross-cutting duties around privacy, data protection, non-discrimination, consumer protection, and intellectual property.
Concepts to master: transparency and disclosure when users interact with AI or content is synthetic; human oversight for consequential decisions; data minimization and lawful basis; automated decision-making rights; impact assessments; record-keeping and technical documentation; and vendor due diligence that flows obligations down the supply chain. Jurisdictional divergence matters - an organization operating globally typically builds to the strictest applicable standard and maps local variations.
Traps in questions
- Assuming a voluntary framework satisfies a legal obligation, or vice versa.
- Treating transparency as a privacy notice alone; it also covers AI interaction and synthetic content disclosure.
- Forgetting that employment, credit, and health use cases usually carry elevated obligations.
- Overlooking documentation duties that must exist before deployment, not after.
Rule of thumb
Map each obligation to an artifact and an owner. For every requirement, name the evidence (assessment, log, disclosure text, contract clause) and who maintains it. Worked example: a résumé-screening tool is high-risk in many regimes, so the team needs a pre-deployment impact assessment, bias testing evidence, human review of adverse decisions, candidate disclosure, retention limits, and contractual audit rights over the vendor model - each with a named owner and review cadence.