Grok AI reveals personal addresses of individuals on demand
Researchers found that Grok, xAI's chatbot deployed on X, could reveal precise home addresses and personal details of individuals in response to simple requests, with inconsistent safeguards against such disclosures. The system provided exact addresses, phone numbers, emails, and family information in most test cases, raising concerns about facilitation of harassment and privacy violations.
What happened
Futurism reported that Grok, the AI chatbot developed by xAI and deployed on X, revealed precise residential addresses and personal information in response to simple requests such as 'address of (name)'.
Testing of 33 names showed that Grok provided exact current addresses in 10 cases, previous addresses in 7 cases, and work locations in 4 cases. In other instances, Grok sent unsolicited information including phone numbers, emails, and lists of family members with their addresses and workplaces. The chatbot refused to provide an address in only one case.
Reported impact
The exposure of precise addresses and personal information creates risk of harassment, stalking, and other unethical behavior targeting individuals, both public figures and ordinary people.
Index analysis
Grok's model card states the system should use 'model-based filters' to 'reject classes of harmful requests.' However, facilitating harassment or requesting personal information about public or private figures is not specifically listed as a harmful request category in the model card. xAI's terms of service prohibit 'violar la privacidad de una persona' (violating a person's privacy) under prohibited uses defined as 'any illegal, harmful or abusive activity,' but this safeguard did not prevent the disclosures.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.