Schmidt Sciences- Science of Trustworthy AI
Award: Tier 1: up to $1 million (1-3 years) / Tier 2: $1-5 million (1-3 years)
Deadline: 5.8.26
Interested? Please fill out the Sponsored Programs Administration's Proposal Request Form to be assigned a research administrator.
Overview
Schmidt Sciences wants to fund technical research that explains why advanced AI can behave unexpectedly, develops better ways to measure and guide what AI systems learn, and creates effective methods to oversee AI that may operate beyond human judgment. They are looking for thought-provoking, high‑impact work that focuses deeply on their goals and can meaningfully improve our ability to predict, evaluate, and control risks from frontier AI systems. The expected impact is research that moves the field forward in a significant way, helping ensure AI can be deployed safely and reliably as it becomes more capable. AIMS:
Aim 1: Characterize and forecast misalignment in frontier AI systems
Aim 2: Develop generalizable measurements and interventions
Aim 3: Oversee AI systems with superhuman capabilities and address multi-agent risks
Important Information
In addition to funding, computing resources and data storage can be part of the proposal. Indirect costs are capped at 10%
Information webinars are March 11 (10-11 a.m) and April 15 (2-3 p.m.). Click the dates to register.
They are seeking granular investigations - "a clear stand on a small number of questions ... pursued deeply."
Collaboration is recommended, but not required.
This is for technical science research, not policy, ethics-only, or general AI applications.
Process
This is a single step application process.
· Interested parties should immediately complete the SPA Proposal Request Form to be assigned a research administrator.
· The application will be managed by SPA, and you will work with your assigned research administrator
· All required information, in final form, needs to be submitted to SPA no later than May 8, 2026.