Penn GSE joins $26M effort to build open AI infrastructure for K-12
Penn's Graduate School of Education is partnering on a $26 million K-12 AI Infrastructure Program to build publicly accessible generative AI tools for schools. Catalyst @ Penn GSE has joined Digital Promise and three core partners to fund grants that help develop AI products grounded in real classroom needs.
Jeremy Roschelle of Digital Promise called the effort ambitious and difficult, noting there's still no standard way to collect education datasets, build benchmarks, or tune models for teaching and learning. The program is built to change that by bringing together teams who can set practical foundations and share them openly.
What the program will fund
The focus is formative assessment-using ongoing feedback from students and educators to fix what current tools miss. The goal: reduce friction, respond to real frustrations, and create AI that actually supports instructional leadership and day-to-day learning.
Grants will help create free, legally unrestricted open datasets for training and evaluating models. These datasets will reflect learner variability and sound teaching principles so tools can respond to different students, subjects, and contexts without baking in bias.
Who's involved
- Catalyst @ Penn GSE
- Digital Promise
- Massive Data Institute at Georgetown University
- Learning Data Insights
- DrivenData
Why this matters for school and district leaders
- Shared, open datasets make it easier to compare tools and hold vendors to evidence-based standards.
- Better benchmarks mean better alignment to pedagogy, not just generic model performance.
- Infrastructure funding reduces duplicative work across districts and can lower long-term costs.
- Focus on formative assessment keeps AI grounded in everyday teaching, not flashy demos.
What educators and students are saying
Graduate students at Penn GSE emphasized the infrastructure-first approach. They stressed that AI is powerful, but it doesn't automatically reflect pedagogy-data must be vetted, validated, and fair.
Students also highlighted the value of cross-organization collaboration. Different perspectives lead to better models because AI learns from behavior and feedback across varied settings, not just one district or product.
Another theme was community. With the right support structure, good ideas can take root, be tested responsibly, and spread where they work.
Penn GSE's recent AI track record
This is the school's second major collaboration with Digital Promise. In 2024, they launched the Data Science Methods for Digital Learning Platforms, a 16-week certificate course, and in November 2024, GSE introduced the Ivy League's first AI-centered education master's in Learning Analytics and AI.
In January 2025, GSE launched the Pioneering AI in School Systems (PASS) program with the School District of Philadelphia. A $1 million Google gift later expanded PASS to five districts across Pennsylvania, Delaware, and New Jersey in October 2025.
What you can do now
- Inventory where formative assessment already lives in your schools and where feedback loops break down.
- Draft data governance rules for classroom AI pilots: consent, student privacy, dataset documentation, and bias checks.
- Update RFP language to require transparent datasets, clear benchmarks, and evidence tied to pedagogy.
- Form a cross-functional working group (curriculum, IT, special education, legal) to evaluate pilots and share findings.
- Invest in staff development so educators know how to interpret model outputs and course-correct in real time.
Further resources
Learn more about Penn GSE's work at Penn GSE. For practical upskilling, browse educator-focused programs at Complete AI Training - Courses by Job.
Your membership also unlocks:
