Home Cyberlaw Ars Technica xAI says an “unauthorized” prompt change caused Grok to focus on “white genocide”