AI Ethics and Responsible AI
Bias, privacy, safety and governance for people who ship AI.
Taught by Asha Karki · AI Product Engineer
What you'll learn
- Explain the main sources of harm in AI systems and the principles used to address them
- Identify where bias enters a machine learning pipeline and measure it with basic fairness metrics
- Apply data protection principles to training data, prompts and model outputs
- Write a model card and a system description that meet transparency expectations
- Design human oversight, safety testing and misuse controls for an AI feature
- Map an AI use case to the EU AI Act risk tiers and outline the resulting obligations
About this course
AI Ethics and Responsible AI is a practical course for product managers, engineers, analysts, policy and compliance staff, and anyone else who helps put AI systems in front of real people. It treats ethics as an engineering and governance discipline rather than a philosophy seminar. You will learn where algorithmic bias comes from and how to test for it with simple fairness metrics, how privacy and data protection apply to training data and prompts, what transparency and explainability mean in practice through model cards and system documentation, how to design human oversight and safety measures against misuse, and the essentials of the EU AI Act and other responsible AI frameworks. No coding background is required, though one lesson includes a short Python fairness check you can run. By the end you will be able to review an AI feature for ethical risk, document it properly and help your organisation build a responsible AI practice.
Course content
7 lessons- 1Why Ethics Is an Engineering ProblemThe kinds of harm AI causes, the principles most frameworks share, and why good intentions are not enough.Preview
- 2Bias and Fairness: Sources and TestsWhere bias enters the pipeline, the main fairness metrics, and a short Python check you can run yourself.
- 3Privacy and Data Protection in AI SystemsData protection principles applied to training data, prompts, outputs and the risks of memorisation and inference.
- 4Transparency, Explainability and DocumentationModel cards, datasheets and system descriptions, plus what explainability means for different audiences.
- 5Safety, Misuse and Human OversightRed-teaming, guardrails, monitoring and designing oversight so that people can intervene when it matters.
- 6The EU AI Act and the Global Regulatory LandscapeRisk tiers, prohibited practices, high-risk obligations, transparency duties and how other jurisdictions compare.
- 7Building a Responsible AI PracticeGovernance roles, review gates, impact assessments and the lightweight processes that make responsible AI routine.