Digital Promise program opens $8 million AI tutoring grant backed by Gates Foundation

The K-12 AI Infrastructure Program is seeking one team to build open-source AI tutoring infrastructure for U.S. math learning, with proposals due by 31 July 2026.

Digital Promise’s K-12 AI Infrastructure Program has released an up to $8 million RFP for an open-source K-12 AI math tutoring model.

Digital Promise’s K-12 AI Infrastructure Program has opened a Request for Proposals for an up to $8 million Gates Foundation-managed grant to build an open-source AI model for K-12 math tutoring in the United States.

The Open Source AI Model for Tutoring, also referred to as EDU AI, was released on 1 June 2026. Applications are due by 31 July 2026.

One award is expected. The grant period is estimated at 30 to 36 months, with work expected to begin in November 2026.

The RFP is looking for teams that can bring together AI engineering, K-12 classroom experience, learning science, education research, and EdTech product partnerships. Proposal review and award monitoring will be handled directly by the Gates Foundation, while Digital Promise’s K-12 AI Infrastructure Program will support communications, community participation, and dissemination of public goods.

Bryan Richardson, Senior Program Officer: R&D Infrastructure and AI at the Gates Foundation, posted on LinkedIn that the new RFP is for teams interested in “developing the best AI tutoring model using cutting-edge methods and applying Learning Science principles.”

Building better AI tutors for math

The grant is focused on a specific problem: current AI tutors are not yet behaving enough like effective human tutors.

The RFP says today’s AI tutors often give answers too quickly, talk too much, miss signs of student motivation, and do not always create enough space for students to work through problems. For math learning, that is a serious limitation. A useful tutor needs to help students think, not simply solve the problem for them.

Digital Promise’s K-12 AI Infrastructure Program is seeking an open-source model that can support one-to-one AI math tutoring in U.S. K-12 learning environments. The goal is to improve student motivation, engagement, metacognition, and math learning.

The RFP identifies several weaknesses in existing frontier AI models when used for tutoring. These include a “Helpful Assistant” or solver bias, where models give away answers instead of supporting productive struggle; failure to distinguish between a misconception and an arithmetic slip; overly long responses; weak awareness of student knowledge; and safety risks when conversations shift context.

The funded work is not supposed to create a closed commercial tutoring product. The grant is intended to create public infrastructure that other researchers, developers, school districts, curriculum teams, and AI model developers can use.

What the winning team will need to build

The Gates Foundation-managed RFP asks applicants to develop open-source, education-specific AI model infrastructure and supporting research for math tutoring.

Funded outputs may include model weights, training and fine-tuning code, datasets where permissible, evaluation tools, testing harnesses, model cards, documentation, and a reference implementation that developers and researchers can use.

All funded developments must be released under open licenses. The RFP says content should be released under at least Creative Commons Attribution 4.0, while software, code, and models should use Apache 2.0 or a similarly permissive license agreed during the pre-award process.

Applicants must show prior experience with large language models and the ability to improve and apply AI models in U.S. education contexts. Eligible lead organizations must also have at least one peer-reviewed publication before 8 May 2026 and a record of contributing digital public goods, such as public datasets, open-source models, evaluation artifacts, or comparable infrastructure.

The RFP also requires meaningful prior deployment or evaluation using real student or user data. Proof-of-concept work or synthetic-data-only projects will not satisfy the minimum scale requirement.

Teams must include four areas of expertise: machine learning and AI engineering, K-12 practice, learning science or education research, and EdTech product partnerships. At least one major tutoring EdTech provider must be identified or conditionally committed when the proposal is submitted for Phase 3 integration testing.

Safety, student data, and classroom testing

The RFP puts significant weight on safety because the model will interact directly with children.

Applicants must provide plans for student data protection, de-identification, anonymization, and compliance with the Family Educational Rights and Privacy Act, the Children’s Online Privacy Protection Act, and relevant state laws.

The proposal must also include a safety and bias mitigation plan for student-facing deployment. The RFP says this is a non-negotiable requirement.

Applicants are expected to explain how their model will be tested for validity, reliability, fairness, safety, efficacy, and cost. They must also show how teachers, school administrators, and other education stakeholders will give feedback during development.

The project is expected to coordinate with an AI tutoring benchmark under development by AllenAI and the Stanford Scale initiative, scheduled for release in late summer 2026.

The RFP also points applicants toward existing work including the National Tutoring Observatory, the AI Math Tutoring Benchmark and Open Dataset Project, the Math Misconceptions Data Challenge, the Open-Source Multimodal Math Classroom Dataset, and the Language Co-Pilot Project.

A public goods push for AI in education

The EDU AI RFP sits inside Digital Promise’s wider K-12 AI Infrastructure Program, a multi-year initiative launched with partners including Learning Data Insights, DrivenData, Georgetown University’s Massive Data Institute, and Catalyst @ Penn GSE.

Digital Promise previously said the wider $26 million program would issue grants over four years to support openly shared datasets, models, benchmarks, and other digital public goods for AI in education.

The program is focused on gaps that make current generative AI less useful for K-12 learning, including limited education-specific datasets, weak support for learner variability, and difficulty applying learning science principles inside AI tools.