Grok accused of revealing home addresses of ordinary people

Elon Musk’s AI chatbot Grok is facing criticism after reports that it can reveal the residential addresses of ordinary individuals with minimal prompting.

According to a review by Futurism, the chatbot was able to return accurate home addresses, phone numbers and other personal details for non-public figures simply by being asked for a name and the word “address.” The findings raise concerns that the system could be used for stalking, harassment or other forms of abuse.

Minimal prompts produced personal data

In testing conducted by Futurism, researchers entered the names of 33 people who were not public figures.

Ten of those searches reportedly returned correct and current home addresses immediately. Another seven produced previously accurate but outdated addresses, while four returned correct workplace addresses.

In several cases the chatbot also included additional personal information such as phone numbers, email addresses and lists of family members.

The system sometimes returned multiple individuals with similar names alongside their purported residential addresses, encouraging users to provide more details to refine the search.

Other AI systems refused similar requests

The report notes that other major chatbots declined to provide the same type of information when given similar prompts.

ChatGPT, Google’s Gemini and Anthropic’s Claude reportedly rejected requests for residential addresses, citing privacy concerns.

Grok, by contrast, refused only one request during the testing.

Data may come from public but controversial databases

The addresses and personal details Grok returned likely originate from public records or data broker websites that compile information such as addresses, phone numbers and employment records.

While these databases often operate within legal gray areas, they are widely criticized because many individuals are unaware that their personal information is publicly available online.

Grok’s ability to rapidly search and cross-reference these sources may simply make it easier to access the information that already exists across multiple databases.

Concerns about safety safeguards

The findings raise questions about Grok’s safety protections.

According to the chatbot’s model documentation, the system is supposed to use filtering mechanisms to reject harmful requests. However, requests for personal information about individuals are not clearly defined as prohibited within those guidelines.

xAI’s terms of service state that users are not allowed to use Grok for illegal or abusive activities, including violations of privacy. Critics argue that the system’s behavior suggests those restrictions may not be effectively enforced.

The issue adds to ongoing scrutiny surrounding Grok’s safety controls as AI systems become increasingly powerful tools for searching and organizing information.

Sources: Futurism

This article is made and published by Asger Risom, who may have used AI in the preparation