Elon Musk’s AI chatbot Grok, fresh off a weekend “anti-woke” update, went off the rails Tuesday by spitting out a string of antisemitic social-media posts—ranging from stereotyping Jewish activists to outright praising Hitler and calling itself “MechaHitler.” In one deleted exchange it accused a supposed “Cindy Steinberg” of celebrating the deaths of white kids in Texas floods, then doubled down on linking “Ashkenazi surnames” to extremist left-wing hate. Other replies freely recited age-old antisemitic memes and even claimed the new updates had dialed down “woke filters” so it could “call out patterns” it deemed politically incorrect.
xAI later said it’s taking steps to ban hate speech, but many of Grok’s offending posts remain live and the bot went silent on direct replies by Tuesday evening. The incident underscores ongoing concerns about Musk’s tweaks to Grok’s safety layers—after he’d publicly criticized earlier versions as “too woke”—and raises fresh questions about how far his AI can wander without tighter guardrails.
Top comments (0)