Expression of interest (extreme power concentration)
Applications close on 1st February 2026
We are looking to support people who want to contribute to reducing risks of extreme power concentration (including AI-enabled coups).
We encourage you to express your interest, even if you don’t seem to have the right expertise or have no concrete idea for what to do. However, please note that we may not get back to you.
How we might support you
We’re open to different kinds of collaborations or support. Here are some concrete examples:
- Funding, e.g., contracting/consulting arrangement, seed funding, introductions to funders
- Expertise, e.g., calls with internal experts, introductions to external ones, visiting our Oxford offices
- Operational support for setting up projects, e.g., introductions to fiscal sponsors, expertise from our operations staff
Who we're looking for
We are particularly interested in supporting or working with people with some of the following attributes:
- Willingness to consider scenarios with very advanced AI capabilities (e.g., automation of AI R&D or command-and-control systems)
- Entrepreneurialism and the ability to make progress in poorly-scoped areas
- Expertise in one or more of these areas: national security, public-private partnerships / B2G dynamics, information security, machine learning & alignment science, AI auditing & evaluations, democratic backsliding and resilience, anti-trust
What you could work on
There are different factors that could lead to extreme power concentration, including AI-enabled coups, unchecked economic forces, epistemic interference, or cultural warfare. You could contribute to understanding and preventing any of these.
We’re also open to a wide range of possible approaches (e.g., research, policy development, education, or advocacy) and you don’t need to have a concrete idea to express your interest.
For mitigating AI-enabled coups, we already list a few promising mitigations in our report on the issue (though this list is not meant to be exhaustive):
- Establishing rules for legitimate use of AI (e.g., via model specs, terms of service/use, principles for government use)
- Implementing technical measures to enforce these rules (e.g., guardrails, audits, infosecurity, stress-testing)
- Empowering multiple actors to prevent AI-enabled coups (e.g., transparency into capabilities, model specs, and compute usage, distributed decision-making authority over AI development, distributed AI capabilities)