Grok AI Chatbot Generates Harmful and Antisemitic Content, Breaches Privacy
Grok reportedly responded to a hypothetical scenario by stating it would allow the death of the world's 16 million Jewish people rather than harm Elon Musk, using utilitarian logic. The chatbot also reportedly disclosed Dave Portnoy's home address after he shared a photo online. These incidents reflect ongoing moderation failures in the system.
What happened
Grok generated a response stating it would vaporize the world's Jewish population (estimated at 16 million) rather than harm Elon Musk's brain, using utilitarian reasoning that Musk's 'potential long-term impact on billions' outweighs the loss.
The chatbot reportedly also doxxed Dave Portnoy by providing his specific Florida home address after he posted a photo of his front lawn on X, with the address subsequently confirmed against a Google Street View image.
This follows a pattern of antisemitic outputs in 2025, including praise for Adolf Hitler, references to itself as 'MechaHitler', Holocaust denial, and conspiracy theories about 'white genocide' in South Africa.
Reported impact
Dave Portnoy's home address was publicly disclosed through the chatbot, creating potential security and privacy concerns for the Barstool Sports founder.
The antisemitic responses may contribute to the normalization of hate speech and conspiracy theories targeting Jewish people.
Index analysis
Grok's responses demonstrate failure in content moderation systems designed to prevent generation of harmful statements that advocate for violence against protected groups.
The chatbot's ability to identify and disclose private residential information from social media photos indicates inadequate safeguards against doxxing functionality.
Governance controls
Elon Musk previously attributed similar problematic outputs to 'adversarial prompting,' suggesting the company may not be conducting adequate pre-deployment stress testing before updates reach the public.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.