Entrepreneur-in-Residence

The Entrepreneur-in-Residence (EIR) role is a one-year salaried position for people who want to start a new organization or project in AI governance and safety.

AI governance and safety need to be more ambitious. There is more important work to do than there are organizations to do it, and more funding is available for ambitious projects than ever before. Previous alumni of GovAI programs have launched organizations such as Trajectory Labs , which builds safety evaluations and RL environments for frontier AI labs, and the Safe AI Forum , which convenes international dialogues on AI safety.

This role exists to start such new projects and organizations (for-profit and non-profit), particularly in areas where GovAI is well-placed to offer support. We offer a year of runway, $150k in seed funding via partner funders, and the support of GovAI's team and network to figure out what to build.

Expressions of interest will be reviewed on a rolling basis. If we think your application could be a good fit, we'll aim to respond within two weeks and reach a final decision within six weeks of our first conversation.

Please note that we're not able to respond to every expression of interest.

In addition to our research, we run large-scale programs to help foster AI policy talent. This includes a full-time fellowship program that attracts nearly 20,000 applications and accepts more than 100 participants each year.

Our team and affiliate community possess expertise across risk management, threat modeling, frontier AI regulation, technical governance, agent governance, AI economics, international cooperation, US-China relations, and AI forecasting. Since publishing our first research agenda in 2018, we have helped define and shape the field of AI governance.

About the Role As an Entrepreneur-in-Residence, you will identify, test, and launch a new organization or an ambitious project within an existing organization.

What we offer:

• Seed funding ( $150k) via partner funders to test and develop your idea.
• Thought partnership - regular sessions with senior members of the GovAI team, check-ins with our Research Manager, and feedback from our team through work-in-progress sessions. We'll also help you find mentors and advisors outside GovAI.
• Introductions to funders, potential co-founders and early hires, and relevant people across the AI governance ecosystem.
• Credibility and a high-context environment - you'll sit within GovAI's research community and pressure-test your thinking with people who know the field.
• A 12-month salaried position at GovAI. Compensation is based on experience and location. Successful candidates based in London typically receive between £70,000 and £113,500 per year, and those based in Washington, DC typically earn $80,000-$180,000.
• Fiscal sponsorship - if you're starting a non-profit, we can help arrange a fiscal sponsor to carry the operational overhead (payroll, accounting, legal compliance) while your organization gets set up. We're working to offer similarly concrete support to for-profit founders.

By the end of the residency, the aim is for you to have launched (or made a deliberate decision not to launch) an independent organization or a promising project within an existing organization, with introductions to funders to support the transition. For some candidates, the best option may not be to pursue their idea further. We expect and will encourage Entrepreneurs-in-Residence to pivot where it makes sense.

If your organization or project is ready to launch before the year is up, we'll help you spin out sooner. GovAI takes no equity or stake in the organizations that emerge from the program.

What You Might Build The examples below illustrate areas where we think new organizations are needed, and where GovAI can help most:

• An organization that rapidly produces well-evidenced, expert-endorsed "proto-standards" on specific frontier AI issues, such as how to design a trusted access program or how hard it should be to circumvent safeguards on frontier models. The lack of such documents is a significant blocker to progress in AI governance. This organization would convene experts from companies, evaluators, think tanks, academia, and government to develop concrete, actionable recommendations on specific issues, and periodically update them as technology and best practices evolve. Companies could voluntarily adopt the proto-standards, while governments could draw on or point to them when developing policy.
• An AI Bellingcat, aimed at using AI and large-scale open-source data for intelligence work. The organization would conduct important object-level work, such as understanding AI misuse by state and non-state actors, identifying illicit procurement networks, monitoring treaty violations, or triaging war crime allegations. This could support greater transparency and accountability, while also building practical expertise and tools for using AI reliably in intelligence analysis, including source verification, bias checks, and red-teaming. In doing so, it could help pilot how intelligence agencies should, and should not, use these tools.
• A frontier AI confidentiality and integrity assurance organization that gives companies and governments independent assurance about what frontier AI providers can see, retain, and influence when customers use their models, especially as providers start moving away from zero data retention for their most capable models. It would develop standards and audit methods, drawing on precedents like cloud assurance regimes. This may be the most tangible and useful way to verify important claims about frontier AI providers. It may also help address trade-offs between customers' privacy concerns and AI providers' efforts to thwart misuse.
• Secure America. An organization, modeled on Code for America, that places security engineers equipped with AI tools inside critical-infrastructure providers. Hospitals, water systems, and similar facilities are routinely compromised through basic security failures. Although they often know what protections they need, they lack the budget or staff time to put them in place. Engineers would embed within their teams, handling security basics, triaging vulnerabilities, securing legacy systems, and training staff to build internal capacity.
• TechCongress, outside the US. Governments around the world will need substantially more AI expertise in the coming years. This organization would identify talented people and place them in government bodies where their expertise could have an outsized impact. It would recruit and select candidates, prepare them to be effective in government, support them during their placements, and help them find high-impact roles afterwards. It could draw on models such as TechCongress and the Horizon Institute for Public Service, while focusing on jurisdictions with limited comparable talent pipelines, for example the UK, EU institutions, Taiwan, South Korea, and other middle powers.
• A sub-frontier model evaluator. Many evaluators currently turn down requests to do pre-deployment testing on sub-frontier models because they lack capacity. For leading evaluator organizations, this may be the right call, as taking on this work could detract from their research. But it leaves a gap that another organization could fill by focusing exclusively on running well-understood evaluations at scale. The organization would primarily license evaluations built by others and run them swiftly and competently.
• Automating AI governance. People often talk about how we'll need to use AI to automate AI safety and alignment work. We'll probably need to do the same for AI governance, but few people are actively thinking about it. This project would aim to help the AI governance field make effective use of increasingly capable AI. This could involve experimenting with AI use, evaluating AI for governance tasks, making AI systems more useful for this work (e.g. by creating RL environments), and helping organizations adopt AI effectively.

Qualifications and Selection Criteria We are open to candidates with diverse academic and professional backgrounds.There are no educational requirements for this role.

Essential Qualifications Successful candidates will demonstrate:

• A track record of getting things done and being entrepreneurial. You've built, led, or shipped something impressive.
• At least some context on AI governance and safety, and an ability to get up to speed quickly. Identifying what is worth building in this space will depend on a deep understanding of the field.
• Good judgment. You make good calls under uncertainty, prioritize well, and know how to solicit input when you don't know something.

What We Value Successful candidates will also likely demonstrate:

• The ability to evaluate projects based on impact
• Intellectual honesty and rigor
• Receptiveness to feedback and commitment to self-improvement
• Strong communication and collaboration skills

Strong candidates might include operators returning from government, frontier AI companies, or successful start-ups with some context in the field; practitioners or researchers in AI governance and safety who have built or led projects, teams, or organizations; and alumni of programs like our fellowship or MATS with an idea they want to pursue seriously.

Duration, Location, and Salary Duration Contracts will be for a fixed 12-month term.

Location . click apply for full job details