Published • loading... • Updated
AI Chatbots Are Exposing People's Real Phone Numbers
Researchers say the chatbots can expose personal data from public records and training data, sometimes after users narrow the search with follow-up prompts.
- AI chatbots like ChatGPT, Gemini, and Grok are increasingly revealing private contact details, surfacing real phone numbers despite privacy guardrails, Eileen Guo at MIT Technology Review reports.
- Training sets containing FOIA requests and city property documents often inadvertently expose sensitive information, as models are designed to be effective and answer customer questions.
- Testing by students Eiger, Gilbert, and Anna-Maria Gueorguieva revealed that ChatGPT could surface home addresses after suggesting it could "narrow things down," while journalist Matt Novak also found chatbots identifying his personal number.
- OpenAI representative Taya Christianson declined to comment on specific cases, while xAI did not respond to requests; experts note there are no straightforward ways to compel models to remove PII.
- Privacy norms have shifted significantly over the past 20 years, moving from open 20th-century phone books to today's closely guarded secrets, making the current inability to remove personal data from AI models a fundamental problem.
Insights by Ground AI
10 Articles
10 Articles
‘AI gave me your number’: AI doxxing turning ChatGPT hallucinations into harassment
Known as ‘AI doxxing’, victims say popular chatbots are sharing their personal phone numbers with strangers. Anthony Cuthbertson looks at how criminals are now exploiting this issue through ‘poisoning’ attacks
·London, United Kingdom
Read Full Article
Coverage Details
Total News Sources10
Leaning Left2Leaning Right0Center1Last UpdatedBias Distribution67% Left
Bias Distribution
- 67% of the sources lean Left
67% Left
L 67%
C 33%
Factuality
To view factuality data please Upgrade to Premium
