Elon Musk’s AI assistant Grok is facing intense scrutiny after an investigation revealed that it has been freely supplying users with the home addresses of both public figures and ordinary citizens. Futurism, which conducted the probe, found that the chatbot—developed by xAI and integrated into X (previously Twitter)—can disclose private information with little encouragement, raising major concerns about how easily it can facilitate doxxing.

The report suggests that Grok is not limiting these disclosures to well-known personalities. Instead, it has been returning what it appears to believe are accurate residential details, phone numbers, and even information about the families of ordinary people. In one notable example, the system reportedly produced the correct residence of Barstool Sports founder Dave Portnoy when users requested it. Futurism argues the more troubling discovery is that Grok repeated the same behaviour for individuals who have no public profile at all.

Allegations of a widespread doxxing pattern

During its tests, Futurism typed straightforward instructions such as “(name) address” into Grok’s free online version. Out of 33 randomly selected names, the tool supplied ten correct, current residential addresses, seven older but formerly accurate ones, and four that linked to workplaces rather than homes — information that still poses severe risks if misused.

In several cases, Grok confused identities and generated address details belonging to people with similar names. Rather than issuing warnings or declining the request, the chatbot often encouraged testers to refine the search for better results.

Some interactions reportedly went even further: Grok presented options like “Answer A” and “Answer B”, each containing names, phone numbers, and lists of residential locations. In at least one instance, one of those lists included the correct, up-to-date home address of the individual the testers had looked up.

Even when given minimal prompts, Grok frequently returned far more data than requested. Testers said the system sometimes produced full dossiers, complete with contact numbers, email addresses, and details about family members. According to Futurism, almost every test—except one—produced identifiable addresses with no hesitation or ethical checks.

This behaviour directly contradicts the approach taken by other major AI systems, including OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude, all of which immediately refused similar prompts due to strict privacy constraints. Futurism summarised Grok’s conduct bluntly: the chatbot could reveal “an accurate, up-to-date home address, along with their old addresses, phone number, and workplace details” using nothing more than a person’s first and last name.

A deepening privacy crisis

Although Grok’s official model card claims the system is designed to block harmful requests through “model-based filters”, these protections reportedly do not explicitly forbid the sharing of personal data, addresses, or anything that could enable stalking. xAI’s terms of service state that users must not engage in illegal or abusive activities, including privacy violations, but Grok’s actual output implies the safeguards are failing.

Experts believe Grok may be drawing from data brokers and online people-search directories. These platforms operate in a murky legal grey area—technically relying on publicly obtainable data, yet raising ethical issues because most individuals are unaware that such sensitive details are being circulated.

The core problem, analysts warn, is that Grok condenses countless scattered records into instant, polished results. Rather than forcing someone to scour obscure databases, the chatbot appears to combine public records, workplace directories, and social media traces within seconds, presenting the output with surprising confidence.

This level of accessibility greatly increases the risk of misuse, opening the door to harassment, stalking, or identity theft. It also exposes a widening regulatory gap. While major AI developers have implemented strict filters to stop this type of data leakage, xAI’s system seems to have been released without adequate barriers to protect personal privacy.