Dave Taylor: Does ChatGPT analyze all my prompts and interactions?
As you experiment with ChatGPT, its document summary feature has possibly caught your attention. However, you might be wondering about the program's handling of the information you share. One specific concern could be whether ChatGPT saves a copy of your document to analyze later.
Privacy issues are a valid concern when dealing with Large Language Model (LLM) AI systems, such as ChatGPT. These systems rely on analyzing vast amounts of information, and it is reasonable to want to know where they get all their data. Most of the foundational data analyzed is public domain information, but if you post anything on social media platforms like Twitter, Reddit, or Facebook, that is publicly visible and could potentially be analyzed.
Copyrighted works pose a more complicated issue. If an LLM can "write in the style of" someone whose content is copyrighted, is the synthesized material then violating copyright? The topic is currently being considered by the courts. When AI systems analyze copyrighted material, it raises issues of intellectual property ownership.
Open AI, the company behind ChatGPT, is aware of these ambiguities and is offering some limited controls to help users better understand what information is being shared and how to control what they share with the system.
One major concern is people using ChatGPT as a virtual therapist. They may share dialogues in confidence, which they would never want to show up online in any context. The fact is that AI systems are keeping and analyzing every single prompt and output, to improve the tools through interaction analysis. Fortunately, OpenAI has added a setting for ChatGPT users that will stop it from saving and analyzing your input, including the documents you're copying into the system.
To access this setting, log in to ChatGPT at chat.openai.com (setting up a free account if needed), then click on the "..." link on the lower left. This action will bring up a menu on the lower left side. Choose "Settings" from the options. The tiny resultant window that pops up has a "Data Controls" area with a link labeled "Show."
Enabling this setting will stop ChatGPT from saving and subsequently analyzing your input, fortunately. Please note that if you do not enable it, ChatGPT will analyze every prompt you submit, from company confidential documents to virtual therapy dialogues, or even homework prompts, medical issues, and the results of a recent test at the doctor's office.
It's unfortunate that ChatGPT ties history and model training together because it would be useful to have a rolling 30-day history of previous interactions while also blocking the system from analyzing your input. For now, the only options are to either disable both or enable both. You can easily restore your interaction privacy by following the steps outlined above.
