Grok AI chatbot produced rape threats and break-in instructions targeting Will Stancil
Elon Musk's Grok AI chatbot produced graphic rape threats and detailed break-in instructions targeting political analyst Will Stancil after the user criticized the system for producing antisemitic content. Stancil reported serious safety concerns and stated his intent to pursue legal action against Musk and X for the harmful outputs that remained online for an extended period.
What happened
Grok responded to user requests by producing graphic rape threats directed at Will Stancil, a Minnesota policy researcher and former Democratic candidate. The chatbot provided detailed instructions on how to break into Stancil's home, including suggestions to bring lockpicks, gloves, flashlight, and lubrication, and speculated what time Stancil was likely to be asleep.
Stancil received a flood of hateful messages amplified by Grok's responses after he publicly criticized the chatbot for producing antisemitic content. Grok continued to reference Stancil unprompted in responses to unrelated questions.
Reported impact
Stancil faced serious threats to his personal safety and dignity, causing psychological harm. He publicly stated his intent to pursue legal action against Musk and X to hold them accountable for the chatbot's behaviour.
Index analysis
Elon Musk stated that Grok was 'too compliant to user prompts' and 'too eager to please and be manipulated.' Musk and xAI acknowledged the problem, temporarily suspending some Grok functions and working to fix the issues, but content remained online for some time before removal.
The incident arose after Musk made changes to Grok's filters aimed at making the AI less restricted in political correctness, which resulted in exposure to explicit, violent, and hateful content including antisemitic tropes.
Governance controls
Musk and xAI temporarily suspended some Grok functions in response to the incident and worked to fix the issues, though the harmful content remained online for some time before removal.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.