COMPASS: AI X-Risk Track

COMPASS (Capacity-Oriented Mentorship for Public Administrators on AI Safety & Strategy): AI X-Risk Track trains sitting Global South government officials to manage AI x-risk, so their governments are part of preventing it.

Endorsed by+2

Project Details

We believe that every government has a role to play in avoiding and managing x-risk from advanced AI. Right now, only a handful of countries are actively working on this issue even though the Global South has just as much stake as others. Global South also needs to build the same capacity to manage these risks.

COMPASS: AI X-Risk Fellowship is a focused track developed as part of the broader COMPASS (Capacity-Oriented Mentorship for Public Administrators on AI Safety & Strategy) initiative. We are developing this main COMPASS Project as a capstone project under the Successif AI Policy Strategy Fellowship. While the aim of broader COMPASS initiative is to build the capacity of civil servants on AI Safety and governance, the AI X-Risk track is focused specifically on catastrophic and x-risks from AI.

It is an 8- week fellowship that will train sitting civil servants and public administrators from the Global South who are already in a position to contribute to policy development, institutional readiness, and strategic decision making, rather than students and early career professionals who may still need to find a job or a funding for their research works before even they can make an impact. Most AI safety fellowships target students and early career professionals who are years away from the decision making roles.

COMPASS AI X-Risk intentionally prioritizes civil servants who already have a job and can use the learnings right away. Our fellowship skips the theory-only and reading-heavy model that most fellowships run on. Instead we include self-paced content along with a live weekend session delivered by a subject matter expert/mentor each week.

We will design the curriculum from the ground up taking into account what the global south officials actually deal with in the real world such as slow bureaucracies, thin resources, competing vendors, and limited leverage in the multilateral rooms.

Fellows will learn what AI x-risk actually is, how advanced general-purpose AI systems work and why their capability trajectories matter, and what governments can do practically including monitoring, red lines, incident response, procurement guardrails, and multilateral negotiations. During the fellowship, each fellow will also explore and develop a risk management plan for their own country, covering catastrophic and x-risk from Ai and how their country's preparedness contributes to global AI Safety. They will also receive feedback from the mentors and peers and join the COMPASS alumni network.

Theory of Impact

Preventing catastrophic outcomes from advanced AI isn't something a handful of governments can manage alone. Every country with a stake in the outcome needs to be able to play its part, monitoring deployments, holding AI companies accountable, and showing up as an informed party in the agreements meant to keep this risk contained. Right now that capacity sits almost entirely with the US, UK, and EU. Most of the world isn't in a position to manage this risk right now. That means they can't monitor it, can't act on it, and any agreement that leaves them out is going to have gaps.

This fellowship closes that gap at the national level, by putting the capacity to manage x-risk into the hands of officials who already sit inside government and can act immediately. We are not claiming that 25 officials will solve global AI x-risk on their own. Our claim is that avoiding it globally requires this capacity to exist everywhere it's currently missing, and this is how we should start building it.

People

Olin Thakur

Team Member

Funding Details

Start Date: Aug 1, 2026
End Date: Oct 31, 2027
Expected Duration: 1 year
Funding Raised to Date: -
Annual Budget: -
Monthly Burn Rate: -
Current Runway: -
Funding Goal: -
Funding Stage: seeking first grant
Fiscal Sponsor: -

Track Record

Techplomat at Techplomacy Foundation, leading work at the intersection of emerging tech/AI, diplomacy, governance, and national security. Editor in Chief at Techplomacy Magazine, an emerging independent, nonpartisan, nonprofit mag for global dialogue on AI safety and governance. More info: LinkedIn
I bring 18+ years of experience delivering humanitarian and development initiatives across 21 countries, managing large-scale operations, and building partnerships across governments, multilateral organizations, and civil society.
Previously, I founded and successfully exited two startups, demonstrating experience building organizations from inception to scale.
I am an AI Safety & Governance fellowship alumnus of BlueDot Impact, AI Safety Collab, European Network for AI Safety, Center for AI Safety, and Successif.

Grants Received– no grants recorded

Add grant

Funding Asks

grantmaking.ai Launch Round

Applied

COMPASS (Capacity-Oriented Mentorship for Public Administrators on AI Safety & Strategy): AI X-Risk Track We train sitting Global South government officials to manage AI x-risk, so their governments are part of preventing it.

Minimum

$45,200

Ideal

$115,500

What this ask is

We believe every government has a role to play in avoiding and managing existential risk from advanced AI. Right now, that job falls to a handful of countries. The Global South has just as much at stake and needs the same capacity to manage that risk.

COMPASS: AI X-Risk Fellowship is a focused track developed as part of the broader COMPASS (Capacity-Oriented Mentorship for Public Administrators on AI Safety & Strategy) initiative under development, a proposed fellowship for civil servants that we are designing as a capstone project under the Successif AI Policy Strategy Fellowship. While the broader COMPASS initiative aims to build government capacity on AI safety and governance, this AI x-risk track focuses specifically on catastrophic and existential risks from frontier AI.

It is an 8-week fellowship that will train sitting civil servants and public administrators from the Global South who are already positioned to contribute to policy development, institutional preparedness, and strategic decision-making, rather than students and early-career professionals who are still years away from holding roles with direct influence over government policy.

Most AI safety fellowships are focused on students and early career professionals who are years from actually being in a position to act. This means even if we train students and researchers, they may need to find a job first or research funding to continue their work and still may not have direct impact. COMPASS does the opposite by training people already inside government, so that what they learn gets used right away. It also skips the theory-only, reading-heavy model most fellowships run on. Every week includes self-paced content with a live weekend session led by someone who has actually worked in this space and not just written about it. And the curriculum won't be a repackaged EU or US framework, but will be built from the ground up around what Global South officials actually deal with such as slow bureaucracies, thin resources, competing vendors, limited leverage in multilateral rooms.

Fellows will learn what AI x-risk actually is, how advanced general-purpose AI systems work and why their capability trajectories matter, and what governments can practically do including monitoring, red lines, incident response, procurement guardrails, and multilateral negotiations. During the fellowship, each fellow will explore, build, and present a real risk management plan for their own country, covering catastrophic and existential risk from AI and how their country's preparedness contributes to global AI safety. They get feedback from mentors and peers and join the COMPASS alumni network.

How the money will be spent

At $45,000 (minimum), one cohort of 25 fellows:

Program Director (part-time): $16,000
Program Coordinator (part-time): $8,000
Curriculum design (AI x-risk content): $7,500
Mentor honoraria (8 sessions, one per week, $700 each): $5,600
Platform and tools: $1,000
Outreach and applications: $2,000
Monitoring and reporting: $1,000
Subtotal: $41,100
Contingency (10%): $4,100
Total: $45,200

At $120,000 (ideal), two cohorts of 25 (50 fellows total):

Program Director: $40,000
Program Coordinator: $20,000
Curriculum design (refined between cohorts): $11,000
Mentor honoraria (16 sessions, 8 per cohort, $700 each): $11,200
Platform and tools: $2,200
Outreach and applications: $6,000
Monitoring, reporting and alumni network coordination: $14,600
Subtotal: $105,000
Contingency (10%): $10,500
Total: $115,500

Discussion

AAbeer Sharma12d(edited)

We need more projects that target major economies that have so far been neglected in AI safety priorities. I’d like to see this project funded and follow its progress live.

Olin Thakur10d

Thanks Abeer for your feedback.

Karolina Gruzel10d

There is at least in my corner of AI safety and governance ecosystem a lot talk about the Global South inclusion, and not only for reasons of equity but also for better outcomes in x-risks reduction. But we need more initiatives like this that provide solutions, and as Onlin outlines focus on action rather than theory.

Olin Thakur9d

Thank you Karolina for the feedback and support.

MMelody ShumEndorsement10d

We need targeted interventions like COMPASS to mitigate x-risks, and co-producing Global South AI safety is essential for that!

Olin Thakur9d

I agree. Thank you Melody for the feedback and support.

Genevieve Shea10d

Equipping public officials in the Global South with the knowledge and tools to engage on AI safety and governance strengthens global resilience. I appreciate the focus on practitioners who can apply these lessons immediately.

Olin Thakur9d

Thanks Genevieve for the feedback. Those who are strongly passionate about this cause area and also may have knowledge but may not have a job opportunity or funding or decision making authority. And those who already have a job and decision making power may not have enough knowledge on the subject matter or guidance on what exactly to do, and network with civil servants from other countries who may be working in same capacity on the same issue. So, after lots of analysis, we believe this could be cost-effective intervention and impactful.

ZZemingEndorsement6d

This fellowship curates the fellowship towards the global south population, raising awareness to prevent the potential AI risks and hazards. Democratizing the understanding of AI risks is democratizing the future awareness and solution, solving data biases, labor market concerns, and X-risks.