Honours / 4th-year · 2026

African Technical AI Safety

University of Cape Town, Department of Mathematics and Applied Mathematics

Assoc. Prof. Jonathan Shock

The 2026 taught route is complete here: Sessions 1–9, then 11–13, followed by a project. A note on these materials: The design, presentation, and content of these pages have been created and enhanced using AI tools (primarily Claude by Anthropic). Materials are checked during preparation, but errors are still possible, particularly in links, citations, or statistics. If you spot anything wrong, please email jonathan.shock@uct.ac.za. See the full AI content disclaimer.

About this course

A technical introduction to AI safety for students who understand neural networks and can program in Python. The 2026 taught route covers Sessions 1–9, then evaluations, technical governance and one worked interpretability case in Sessions 11–13, followed by a research project. Session 10 is skipped; the numbering follows the original course plan.

What makes it African technical AI safety is a through-line rather than a single week: frontier models are measurably less safe in African languages (a robustness and evaluation problem); compute, energy and data sovereignty shape who can build and govern these systems; and relational ethics and the present-harms debate ask, throughout, whose safety, whose risks, whose values? The labs and project are built to run on Google Colab.

The development of this course was supported by the International Development Research Centre (IDRC) and the UK Foreign, Commonwealth & Development Office (FCDO) through the African Hub on AI Safety, Peace and Security. This work was carried out with financial support from UK International Development and the International Development Research Centre, Ottawa, Canada as part of the AI for Development: Responsible AI, Empowering People Program (AI4D). The views expressed herein do not necessarily represent those of IDRC or its Board of Governors, or the UK government’s official policies.

Read more about the course, the convenor, funding, licensing, and how to cite →

Part I · Foundations and the safety problem
1

Orientation and foundations

2

The alignment problem and the physical substrate

Part II · How models are trained and aligned
3

LLM training, end to end

4

Scalable alignment methods

Part III · Robustness, evaluation and governance
5

Robustness and adversarial ML

6

Evaluations and technical governance

Part IV · Interpretability and project
7

One circuit, then project work

After Session 9

The 2026 taught route gives the sequence, bounded pre-class reading and project starting points in one place.