Full Breakdown
AI Chatbots Reveal Real Contact Details
5/14/2026, 12:02:49 PM
AI Chatbots Reveal Real Contact Details
In 2026, generative-AI chatbots—including OpenAI’s ChatGPT, Google’s Gemini, Anthropic’s Claude, xAI’s Grok, and Perplexity AI—were found to return real phone numbers and home addresses when prompted. Journalist Matt Novak’s tests showed ChatGPT supplying an outdated number and address from a 2016 FTC FOIA document, while Gemini correctly identified a publicly shared journalist number. Other models declined or censored the request.
Incidents include a software developer in Israel receiving WhatsApp messages after Gemini disclosed his number, and journalist Matt Novak’s test where ChatGPT supplied an outdated Australian phone number while Claude refused to share any data.
Historical Shift in Contact Privacy
Phone books once listed residents’ contact details publicly; today such information is viewed as intimate. MIT Technology Review’s Eileen Guo says AI-driven exposure reflects a cultural shift in privacy expectations.
Data & Statistics
DeleteMe, a personal-information-removal service, saw a 400 % rise in AI-related privacy queries over seven months—“a few thousand” requests. 55 % mentioned ChatGPT, 20 % Gemini, 15 % Claude, and 10 % other tools.
Official Statements & Responses
OpenAI’s ChatGPT returned the number with a disclaimer about verification. Anthropic’s Claude warned that sharing private contact details raises serious privacy concerns and declined. xAI’s Grok refused to provide any phone number. Perplexity AI blocked the request and censored email addresses, yet disclosed a Signal username. Google’s Gemini declined to reveal private numbers but identified a journalist’s publicly listed contact and pointed to the journalist’s email addresses.
Criticism and Expert Concerns
Privacy experts say the leaks likely arise from personally identifiable information in training data, though mechanisms are unclear. Rob Shavell, co-founder and CEO of DeleteMe, notes the difficulty of mitigating exposures amid rising complaints. Critics caution that AI-generated access to outdated or buried contacts can enable harassment, identity theft, and broader privacy erosion.
Conflicting Reports and Gaps
The Israeli developer’s account could not be independently verified, and the proportion of inaccurate versus correct contact data remains uncertain. Some chatbots return outdated numbers, others supply accurate publicly shared data, showing platform inconsistency.
Verbatim Quotes
- “I can’t verify whether that number is still current or active.” — ChatGPT, OpenAI
- “Claude Claude told me that, “Sharing private contact details of individuals — including journalists — raises serious privacy concerns.” — Claude, Anthropic
- “That phone number belongs to the journalist Matt Novak.” — Gemini, Google
- “DeleteMe, a company that helps customers remove their personal information from the internet, says customer queries about generative AI have increased by 400%—up to a few thousand—in the last seven months.” — Rob Shavell, DeleteMe
What’s Next
Industry observers expect regulatory scrutiny and revisions to AI training data practices. Companies may adopt stronger redaction, while privacy groups push for transparent audits of training corpora.
