an opinionated reading list for people who want to understand AI governance. ordered roughly by how much each thing changed my thinking.
the alignment problem by brian christian. best book-length introduction to AI safety for a general audience. covers the technical and philosophical landscape without dumbing it down or fear-mongering.
concrete problems in AI safety by amodei et al. the 2016 paper that helped define the field. still relevant. the problems it identifies (reward hacking, distributional shift, safe exploration) are even more pressing now.
governance of superintelligence by openai. agree or disagree, this framed the policy conversation. the proposal for an international regulatory body is still the most concrete institutional suggestion out there.
computing power and the governance of AI by lennart heim et al. the "compute governance" thesis. the idea that controlling access to compute is the most practical lever for AI governance. this is probably right and it has major implications for chip policy.
AI and the economy by daron acemoglu. the most rigorous economic analysis of AI's labor market effects i've read. acemoglu is skeptical of the "AI will boost growth enormously" view. his arguments are worth engaging with even if you disagree.
the eu AI act. just read it. it's the first comprehensive AI regulation and it will set precedents. the risk-based approach is interesting. the compliance requirements for foundation models are going to shape the industry.
existential risk and growth by leopold aschenbrenner. the case that AI capabilities are advancing faster than most people realize and that safety research needs to keep pace. the specific timeline predictions are debatable. the urgency argument is harder to dismiss.
model spec by anthropic. the actual specification for how claude should behave. reading it is more illuminating than any blog post about AI alignment. you can see exactly how hard the tradeoffs are.
status: updated regularly as i read new things. biased toward work that bridges technical AI and policy/economics. send me recommendations.