Science of Trustworthy AI

Science of Trustworthy AI RFP

Supports trustworthy artificial intelligence research that improves safety, transparency, and reliability in practical systems.

⚠ This may reflect a past cycle — verify the current call on the funder's site.

The 2026 Science of Trustworthy AI RFP, issued by Schmidt Sciences, invited individual researchers, research teams, universities, national laboratories, institutes, and nonprofits globally to propose foundational technical research on AI safety. The RFP defined three core research aims: characterizing and forecasting misalignment in frontier AI systems; developing generalizable measurements and interventions for AI safety; and addressing oversight challenges under capability gaps and in multi-agent systems including superhuman capability oversight scenarios. The deadline was May 17, 2026 at 11:59 PM Anywhere on Earth; the landing page at schmidtsciences.org was taken down immediately after the deadline closed, consistent with Schmidt Sciences' documented practice for RFPs.

Funding was structured in two tiers: Tier 1 offered up to $1,000,000 for focused investigations, and Tier 2 offered $1,000,000 to $5,000,000 or more for larger multi-year collaborative efforts. All applicants faced a 10% cap on indirect costs regardless of tier. Beyond cash awards, awardees receive access to compute resources, software engineering support through Schmidt Sciences' Virtual Institute for Scientific Software, and API credits with frontier model providers at no additional cost. For-profit entities were not eligible; the program was open to individual researchers, academic institutions, national labs, institutes, and nonprofits without restriction by country.

Applications were submitted exclusively through the Schmidt Sciences smapply.io portal. The accompanying research agenda — which remains publicly available — outlines the scientific priorities motivating the RFP and serves as the primary guide for assessing fit. Schmidt Sciences follows a roughly annual RFP cycle for its Trustworthy AI program; a 2027 cycle would be expected to launch in early 2027. Institutions planning future submissions should track the Schmidt Sciences opportunities portal and allow time for internal approval of the 10% indirect-cost cap before submission deadlines.