Unexpected Encounter with Sensitive Information: A ChatGPT Dilemma

Recently, I had a rather alarming experience while using ChatGPT that raised significant concerns about data privacy and the implications of Artificial Intelligence in our daily lives. What began as an innocent inquiry about the appropriate type of sandpaper to use unexpectedly spiraled into what appears to be a breach of confidentiality involving someone else’s sensitive medical information.

To my shock, the AI didn’t provide a simple answer to my question. Instead, it generated an overview of a drug test belonging to an individual who lives miles away from me. Upon further exploration, I discovered that this response included personal signatures and detailed information that would typically be considered private.

As you can imagine, this left me feeling unnerved. My first instinct was to refrain from sharing the details of the interaction, as I didn’t want to inadvertently distribute any personal information belonging to another person. However, the situation continued to gnaw at me, prompting me to seek advice and share my predicament with others.

In an effort to provide some insight, I did share a portion of the transcript while deleting identifiable sections. Notably, I asked the AI, “What information do you know about me?”—a question I later regretted, as it yielded information about me that I would prefer to keep private. Although I am aware that AI can sometimes produce misleading responses (what’s often referred to as “hallucination”), I cross-referenced the names present in the output, and they appeared to align with their respective locations.

For those interested, my instance of ChatGPT identified itself as Atlas, which is why I referenced that name in my discussions.

I understand that this experience may raise eyebrows, especially as I’m somewhat of a casual user of Reddit and not accustomed to airing my concerns publicly. But I believe discussing these issues is vital as we navigate increasingly complex interactions with technology.

For anyone who wishes to dive deeper into the specifics of this situation, I have linked to my original comment where I detailed my revelations. The discussion surrounding it has opened up conversations about privacy and ethical AI use, and it seems to have resonated with many.

Here’s the link to my comment for those who are curious: Link to the comment

As we