Aligned with the People, Planet, and Progress Sutras, this session explores how evaluations and open source software can help AI for social good scale responsibly. It will examine how red teaming surfaces factuality issues, language gaps, bias, and hallucinations in high-stakes domains including education, food security, and public health. Using the Summit's framing on inclusion, trust, and access, we will gather user needs to guide our open source red teaming app release.