We are a 501(c)(3) nonprofit. You can use this search tool to view our determination and annual returns. Email info@forecastingresearch.org to request any additional relevant public access files electronically.
Funders
Our research is primarily funded by donations. Every contribution supports our work running large-scale forecasting studies to inform decision-making on high-stakes issues. To discuss ideas for forecasting projects or learn more about our specific funding needs, please contact us at info@forecastingresearch.org.
Below is a list of grantmaking organizations and individuals who have contributed $10,000 USD or more to FRI since its founding in October 2022. In addition to the grants below, our first project, the Existential Risk Persuasion Tournament (XPT), was made possible by a grant from the Musk Foundation before the Forecasting Research Institute was established.
We collaborate on impactful research and receive funding to help top organizations prepare well for developments and risks in their domain. We are often restricted from sharing details of collaborations until the research is public. Some collaborations may not be listed below.
Active Site (formerly Panoplia Laboratories)
Date
Collaboration
August 2025 – February 2026
FRI worked with Active Site (a nonprofit that evaluates AI-driven biology, formerly known as Panoplia Laboratories) to forecast the results of one of their experiments and how these might relate to real-world risk. The results can be seen in this post. This project was partially funded by METR.
Anthropic PBC
Date
Collaboration
December 2025 — August 2026
Forecasting the Impacts of ASL-3 Safeguards on Biosecurity Risks was produced as part of an internal project funded by and completed for Anthropic PBC. Anthropic provided input and advice in the development of the report, but all findings and claims represent the views of the authors.
Thinking Machines Lab
Date
Collaboration
June 2026 – October 2026
A private evaluation of the forecasting accuracy of several models at multiple training checkpoints was conducted for Thinking Machines Lab. The work included an additional question set and private leaderboards indicating how the models would have ranked had they been submitted to ForecastBench publicly.