OpenAI alignment lead Boaz Barak to teach Harvard AI Safety course this fall

zainhas · x · 2026-09-06

Boaz Barak, OpenAI alignment team member and Harvard professor, will teach CS 2881R "AI Safety", a graduate course starting September 3, 2026, with the full syllabus now published.

Topics include RSI and AI trajectories, economic impact of AI, RL for post-training, model policies, open-source models, interpretability, alignment in the age of RSI, and AI biosecurity, covering both technical and societal dimensions.

Admission requires Homework Zero (due Aug. 5, 2026) plus math maturity and undergrad ML foundations; in-person attendance is mandatory. The course also assigns two readings imagining AI going wrong by 2030.

Original post →

More from Companies & People

Companies & People channel →