UNESCO – Red Teaming Artificial Intelligence for Social Good: The Playbook, 2025

The document focuses on using Red Teaming as a tool for testing and evaluating artificial intelligence (AI) systems for social good.

It emphasizes identifying stereotypes, biases, and potential harms associated with generative AI. Key findings include the ability of Red Teaming to uncover vulnerabilities in AI models that may contribute to technology-facilitated gender-based violence. The document provides practical examples and recommendations for using Red Teaming to address these issues. It highlights the importance of involving organizations and communities in testing and evaluating AI to prevent negative consequences.