Project summary

Funding shortages for Global South researchers are well known. But even when the money is available, its movement from donor to researcher is often opaque, slow and administratively difficult.

Our non-profit, Exception Raised, in collaboration with Networks for Humanity, is piloting a new approach: $2K digital travel vouchers locked to conference related expenses and redeemable directly at point of sale. Researchers never need to front conference costs out of pocket, submit reimbursement paperwork, or receive funds into their personal bank accounts.

This campaign funds three Indian AI safety researchers travelling to ICML 2026, generating proof that it could be scaled across the country and potentially across the Global South. Fixing research funding bottlenecks is the first step towards attracting serious capital for AI safety work outside the usual hubs.

What are this project's goals? How will you achieve them?

Our objective is to demonstrate that research funding infrastructure can be radically simplified.

Taking the example of travel grants, researchers currently face a series of issues: Researchers earning monthly stipends of $200-$960 are expected to front $2K+ in conference costs out of pocket, sometimes with no guarantee of securing a grant for reimbursement. The process itself consumes significant researcher time in documentation (see below). Further, independent researchers and students are excluded from receiving travel funding support in this system entirely.

We have reimagined researcher travel funding through a programmable voucher: Each voucher is set for specific expense, a defined date range or geography. The researcher redeems it directly with the vendor at point of sale; the vendor settles with Exception Raised on the back end. Funds never have to pass through the researcher's account.

This removes friction on both sides.

For the researcher:

• No upfront payment needed as expenses are deducted from the voucher at point of sale

• No proof-of-expense reimbursement submissions as every transaction is automatically tracked

• No tax or foreign contribution compliance burden, since funds never enter the researcher's bank account

For the donor:

• Every transaction creates a fully auditable trail, simplifying reporting and reducing fraud risk

• Any unutilized balance of the voucher can be redirected to another researcher or returned

This pilot is in partnership with Networks for Humanity. It is a global non-profit organization supported by the Gates Foundation, Google.org and other philanthropic organizations, focused on building networks for open finance and open commerce.

NFH’s patron members include architects of India’s Digital Public Infrastructure (DPI) systems such as Aadhaar (Identity System), UPI (Payment Infrastructure), and DigiLocker (Digital Wallet Infrastructure), along with economists, and former central bank governor of a country. For context, UPI currently processes 650 Million transactions a day, outpacing even Visa.

The program is designed on Vouch.Finance, an initiative of Networks for Humanity, which is building a new approach for delivering social, financial, and purpose-driven value through open finance networks and shared digital infrastructure.

How will this funding be used?

This funding will support three $2,000 travel vouchers for Indian AI safety researchers attending ICML 2026.

We are prioritizing AI safety for the following reasons. Our estimates suggest roughly 20% of India's A* conference publications (leading venues) are regarding AI safety or related topics. Grassroots and community-building efforts have expanded the pipeline of researchers in this area. Lastly, Indians and the Global South are among those significantly affected by AI safety decisions, often with the least say in how they're made. Getting them to global conferences like ICML is one of the most direct ways to close that gap.

We have identified two researchers that we would like to support and are in the process of identifying the third. After a detailed interview process, we have chosen the following AI safety researchers from India accepted at ICML main track:  

·  Kunvar Thaman, solo independent researcher from Chandigarh. Paper: “Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use.

·  Nivya Talokar, independent researcher based in Bangalore, part of a research collaboration. Paper: “Helpful to a Fault: Measuring Illicit Assistance in Multi-Turn, Multilingual LLM Agents”

The voucher will cover major costs heads including flights, hotel, visa, registration, local transport and food. We will use the live pilot to evaluate the following:

• Merchant integration: Whether the voucher system is able to support the full range of conference travel expenses

• Researcher experience: Ease of use and friction points for grantees redeeming funds

• Programmable controls: Ability to set rules governing how and where vouchers can be spent

• Payment tracking: Reliability of real-time transaction visibility and reporting

Who is on your team? What's your track record on similar projects?

• Namrata Rajagopal: Former Project Leader at Boston Consulting Group India focused on emerging technologies. Track-record of enabling innovative financing methods for resource-constrained environments. Published in Harvard Business Review & Fortune.

• Sheetal Chauhan: Founding team & Product Lead (Models) at Sarvam (India’s sovereign AI platform), Senior PM at Microsoft and South Park Commons Fellow

• Dr. Krishnan Thyagarajan: Cross-disciplinary physicist bridging Quantum Computing & AI, Former Associate Director & Senior Scientist at Xerox-PARC. Education and research foundations built at Caltech, EPFL, SRI International, and experience across numerous startups
  

What are the most likely causes and outcomes if this project fails?

A negative result is useful since it tells us if any additional features or operational changes are needed before we run this system at scale. We keep the following scenarios in mind:

·       The selected researchers could drop-off from the pilot due to visa denial or personal circumstances. We will then redirect the voucher to another eligible AI safety researcher in India.

·      While the voucher should theoretically have the same coverage as a card network system, it is possible that pilot researchers in the wild could face difficulty redeeming the voucher for all their expenses. We would have to offer fallback reimbursement option for such scenarios.

·       Our pilot is limited in scale with just 3 researchers in a single conference. Even If we don’t face significant challenges during this deployment, we bear in mind that our sample size is restricted.

How much money have you raised in the last 12 months, and from where?

Exception Raised is a new non-profit with the goal of supporting exceptional researchers found in unexpected places. Prior to entity formation, we ran a pilot funded by Emergent Ventures and an Indian angel investor for the total amount of $25K. This was disbursed among 10 AI researchers in India as a fast small-ticket size grant. The outcomes were: 9 remained actively engaged in research and 5 reported increased ambition in scope. 1 paper selected at ICML main-track, 2 more in progress, 3 open-source projects, 1 potential patent in preparation.