Keiji Kanazawa

This talk introduces the AI Red Teaming Agent, a tool developed within Azure AI Foundry to help AI engineers proactively identify and mitigate risks in their AI systems. It emphasizes that building trustworthy AI is a collaborative effort, akin to building bridges and dams, and that AI engineering requires a systematic approach to testing and iteration. The agent provides a practical way for developers to simulate attacks and evaluate their models' vulnerabilities.