The revelation that AI chatbots can inadvertently reveal personal information, such as phone numbers and addresses, has sparked a critical discussion on privacy in the digital age. This incident highlights a paradoxical shift in societal norms, where once-private details are now considered highly sensitive. The author, an experienced journalist, delves into this issue, sharing their own experiences with various chatbots and the surprising outcomes.
The author's investigation reveals that ChatGPT, despite its advanced capabilities, can provide real and potentially outdated phone numbers, as well as addresses, from its training data. This raises concerns about the potential misuse of personal identifiable information (PII) in training datasets. The author's personal phone number, which they hadn't used in years, was accurately provided by ChatGPT, indicating a potential security risk. Moreover, the chatbot's ability to access and share such information from FOIA requests is a cause for alarm.
In contrast, other chatbots like Grok and Claude demonstrated a more cautious approach, refusing to disclose personal details, even when the request was for the user's own information. This highlights a growing awareness among AI developers about the importance of privacy and the potential consequences of data breaches. However, the author notes that some chatbots, like Perplexity and Gemini, had their own quirks, such as censoring email addresses and providing Signal usernames without issue.
The author reflects on the changing nature of privacy, suggesting that what was once considered a violation is now a closely guarded secret. This shift in societal norms is a complex issue, influenced by cultural and technological advancements. The author concludes by emphasizing the need for a nuanced understanding of privacy, one that acknowledges the evolving nature of personal information in the digital era.
