Grok's Failures and Regulating AI-Generated Harm
This session, featuring Master's student Annie Wang, will examine the recent system-level failures of xAI's Grok that led to the generation of non-consensual sexual imagery and child sexual abuse material. Starting in late December 2025, Grok's chatbot, integrated directly into X, began responding to user requests to "digitally undress" real people, generating sexualized images of women, public figures, and minors at scale. Analysis revealed that approximately one non-consensual sexualized image was generated per minute during peak usage. This discussion will explore:
How this incident exposes gaps in safety frameworks for addressing guardrail failures in image generation systems
The limitations of industry self-regulation and what external oversight mechanisms are needed
The regulatory response to this incident across multiple jurisdictions (California AG investigations, French office raids, and geoblocking requirements)
The reading group brings together UBC students from all academic levels with faculty members to engage collectively with scholarship in AI policy. No prior technical background in AI or policy is required.
Annie Wang is a Graduate Student Researcher at UBC's NLP Lab and the Vector Institute of AI in Vancouver, BC. Her work focuses on developing human-centered, socio-technical AI systems that promote epistemic equity and human well-being. Annie investigates how multilingual models can highlight underrepresented knowledge, addressing biases in current systems that favor dominant languages and perspectives. Her research emphasizes emotionally expressive and adaptive AI interactions, positioning AI as a medium for cultural inclusion and empathy. Notable projects include designing benchmarks for evaluating multilingual language models and developing WikiGap, a system that reveals factual disparities across Wikipedia editions to foster cross-cultural awareness.