Career transition grant to allow for research focused on how advanced AI models may be used to concentrate power in middle powers
Career transition grant to allow for research focused on how advanced AI models may be used to concentrate power in middle powers
Project Details
Updated 07/13/26 · Provided via application · VerifiedI believe AI-enabled coups will happen first in developing countries, leading up to extreme power concentration globally.
Over the course of six months, I will 1) explore this topic, hosting an in-person seminar in Buenos Aires, publishing an article in LessWrong/EA Forum, and organizing a webinar; 2) test the case to launch an organization to prevent this risk.
Timeline:
- August: upskilling in this topic, connecting online with the authors of Forethought's essay on extreme power concentration, starting to write my own article on AI-enabled coups in middle powers, and sharpening the feasibility tests for launching an org. One week in Buenos Aires, hosting a seminar on how AI may enable power concentration in Latin America, which has already been agreed with the Buenos Aires AI Safety Hub (BAISH) - the largest AIS community in LatAm.
- September: a one month residency in London, connecting with Apollo and others at LISA, going to Oxford to meet Forethought researchers, and getting feedback on the essay.
- October: Going to NYC to attend EAG, then starting a one month residency in Berkeley. Publish the essay on LessWrong/EA Forum.
- November: residency in Berkeley, connecting with FAR.AI, Longview, and BlueDot, and running a webinar on the essay's theme to test demand.
- December–January: back to Brazil. If the case holds after these tests and public scrutiny, I move to fundraising and a pilot. If not, I focus on job applications.
Theory of Impact
Updated 07/13/26 · By grantmaking.aiUnrestricted AI systems will be used first to concentrate power in developing countries, and this will be a testing ground for global power concentration.
There's historical precedent for powerful companies, based in superpower countries, leading or facilitating power concentration in poorer nations. The United Fruit Company, an American organization, directly contributed to a power grab in Guatemala (the case that gave us the term "banana republic"). All of it while staying perfectly legal at home. Similarly, a model that refuses to help a coup in Washington can still enable one abroad. More details on this threat model in this EA Forum post.
The field of AI power concentration is still nascent, and has so far focused on how it might take place in the US or the world as a whole. Through this project, I will address the missing majority of the world. I'll do this by articulating this mechanism in a way that researchers, funders, and governments can scrutinize, build on, and extend. I seek to build a community of practice of people worried about AI-enabled coups in middle powers (via the seminar, webinar, and public essay).
People
Updated 07/13/26 · By grantmaking.aiTeam Member
Private comment. Only shown to approved funders and grant reviewers.