What To Do When AI Fails? AI Incident Response

Like all powerful commercial technologies, AI can fail. AI failures might consist of discriminatory behavior, privacy violations, or even security breaches that can lead to lawsuits, regulatory fines, or worse. So what can organizations do to avoid these pitfalls? This talk reviews public AI incidents to illustrate common AI failure modes. It introduces a new approach to “incident response” tailored specifically to AI. And it presents resources to help organizations learn from past failures and get ahead of future incidents: tabletop questions, the AI incident database, other public AI incident repositories, and an AI incident response plan. Participants will leave understanding how AI creates risk for the organizations that employ it, how organizations should plan for AI incidents, and how to react decisively if their AI causes problems.