Personal data: what AI models can reveal

As the digital publication t3n reports, the rise of generative AI is putting increasing pressure on data protection. Telephone numbers, home addresses and internal company information are now easier to access than they once were. While widely used AI models do sometimes block requests for personal information, there are still cases in which they reveal more than they should – or are legally permitted to.

Users on Reddit have reported, for example, that Google Gemini provides private telephone numbers as supposed customer service hotlines. The origin of these numbers remains unclear. Conversely, security experts warn that chatbots sometimes recommend manipulated support numbers that were deliberately published for fraudulent purposes.

Even accurate personal data is sometimes exposed. Although such information often comes from publicly available sources, AI systems can bring it together and make it considerably easier to retrieve.

However, the privacy risks vary between AI models. While ChatGPT, for example, often states that it will not provide sensitive information such as private telephone numbers or exact addresses, t3n describes Grok in particular as taking a more permissive approach to privacy.

By Maria Kleiner

Source: Digital Pioneers, 16 May 2026, by Tobias Weidemann