Course notes

These notes accumulate into the CS120 course reader. Each note has a stable page; this contents page lists only published notes, grouped by module in meeting order.

Module 1 — Framing safety and assurance

Framing Safety and Assurance2026-09-22
What is the Object of Assurance?2026-09-24
What Counts as an Assurance Claim?2026-09-29

Module 2 — Reactive model behavior and development-time optimization

From Data and Objectives to Model Behavior2026-10-01
Proxies and Model-Selection Pressure2026-10-06
Robustness Beyond the Development Distribution2026-10-08
Data as Specification2026-10-13

Module 3 — Sequential decision-making and agency

From Prediction to Action2026-10-15
Reward hacking and control-loop failures2026-10-20
Goal-directedness and long horizons2026-10-22

Module 4 — Adaptive interaction and sociotechnical dynamics

Human feedback, preferences, and reward models2026-10-27

Citation

If you use the materials or repository here is citation information:

MLA

Robertson, Zachary. CS120: Introduction to AI Safety. Stanford University, Autumn 2026, https://stanford-cs120.github.io/fall2026/.

BibTeX

@misc{robertson2026cs120,
  author = {Robertson, Zachary},
  title = {CS120: Introduction to AI Safety},
  howpublished = {Stanford University},
  year = {2026},
  url = {https://stanford-cs120.github.io/fall2026/},
  note = {Autumn 2026 course materials}
}