PolicyA weekendGuided
Run an AI policy tabletop exercise
Gather 6–12 friends, give each a stakeholder role (lab, regulator, journalist, public) and play out how society responds to a fictional AI incident.
PeopleReading & writingView project
Governance1–3 hoursStarterFree
Host a one-hour AI safety discussion
Pick one short reading, invite 4–8 people, and run a focused one-hour discussion with three good questions.
Download PDFPeople
Why it matters
Many people in AI safety got involved through a friend or a reading group, and hosting one is something a beginner can do straight away.
Output: A run session and a one-page guide others can reuse.
Steps
0 of 5 done-
Choose a short reading
20 minUnder 20 minutes to read. One of the “idea lists” or explainers on a theme page works well.
-
Write three questions
20 minOne to check understanding, one that splits the room, one about what to do next.
-
Invite people
20 minAsk 4–8 people personally. A personal invite beats a poster.
-
Run it
1 hrTen minutes of silent reading if people have not read it, then your three questions. Keep it friendly and welcome basic questions.
-
Close the loop
15 minShare the reading and suggest a project each person could try next.
All steps done
Share what you made, then try another project. The PDF version has room for notes.
Related projects
Philosophy1–3 hoursGuided
Explain one safety idea to a friend
Write a short, jargon-free explainer of one AI safety idea (reward hacking, sycophancy, compute governance, DNA screening) for someone outside tech.
Reading & writingView project
PhilosophyA few weeksStarter
Run your own collective constitution
In Anthropic and CIP’s study, about 1,000 Americans used the open-source tool Polis to write a constitution for an AI. Run the same process with 30–100 students and compare your result with theirs.
PeopleReading & writingView project
Further reading
- Requests for discussion on AI infrastructure security and verification (opens in a new tab)Idea listOpen questions on verification.
- MATS AI Safety Strategy Curriculum (opens in a new tab)CourseAs part of their program, MATS runs a series of discussion groups focused on topics relevant to prioritizing research in AI safety. This is the curriculum.
- Synthesis of the state of technical AI verification (opens in a new tab)Reading10 min (hours if you skim the linked works)
- BlueDot: Frontier AI Governance course (opens in a new tab)CourseFree 25-hour cohort course.
- Lennart Heim’s blog (opens in a new tab)Idea listExamples of short, focused analyses.
Related mentored projects
- Developing Updated University Group Fellowship Curricula (opens in a new tab)SPARNixon Hanna · Generator Residency, MIT AI Alignment
- Mitigating Intentional Loss of Control Risk Through Interoperability Standards for Agentic AI (opens in a new tab)SPARKevin Kohler · Simon Institute, UN University, AGI Preparedness Institute
- Developing economies in the post-AGI era (opens in a new tab)SPARAndrei Potlogea · University of Edinburgh