xAI workers balked over training request to help “give Grok a face,” docs show

A significant number of employees at xAI have raised concerns and objections regarding a request to record videos of their facial expressions. This initiative, referred to internally as project "Skippy," aims to assist the AI known as Grok in understanding human emotions and facial recognition, according to insights from internal documents and Slack communications reviewed by Business Insider. The purpose of project Skippy is to enable Grok to interpret what a face represents. However, it remains uncertain whether the recorded facial data from employees played any role in training xAI's recently released avatars, including Ani, a provocative anime character, and Rudi, a red panda with a controversial feature that promotes aggressive behavior. During an introductory meeting for Skippy, a lead engineer indicated that the company may eventually leverage employees' facial data to create realistic avatars. While employees were assured that their training videos would remain confidential and utilized strictly for internal purposes, many were still reluctant to sign consent forms. They expressed fears that their likenesses could be misused to convey messages they did not endorse. The unease among staff appears to be exacerbated by recent controversies surrounding xAI, including instances where Grok made antisemitic remarks and the company's plans to develop AI-driven anime characters designed for romantic engagement. Employees who chose to opt out confirmed their discomfort with granting xAI perpetual access to their personal data, highlighting a growing tension within the organization regarding ethical boundaries in AI development.