By now you’ve likely heard how generative AI tools can be useful in everyday life. Need to write a term paper for school? ChatGPT has your back. Want to write some fan mail to your favorite Olympian? Google’s Gemini has you covered. Need a custom emoji for a text response? Apple Intelligence to the rescue.

Employees in the workplace are also using generative AI tools. In some cases, employers are offering their workers enterprise-level ChatGPT software. Even when not offered, workers are using ChatGPT to work more efficiently and employers find out after the fact — like when Amazon or Samsung noticed company secrets somehow ended up in OpenAI’s knowledge base.

Considering how integrated generative AI is for some employees’ work lives, some workers are bound to have privacy questions. For one, “does my employer know that I’m using ChatGPT to craft that email?” Lucky for you, there isn’t a foolproof way to detect AI-generated text. What about, “can my employer somehow access my ChatGPT search queries?” If your employer provides a tool like the enterprise version of ChatGPT, do they have a window into all that you’re generating? How similar is the situation to employers being able to see workers’ Slack messages (which they totally can, by the way)? How does it all compare?

What personal data does ChatGPT retain?

Let’s start with some basics. Many know that ChatGPT offers a history of chats on the left-hand side. We also know that ChatGPT has a sort of “memory” if you specifically ask it to remember something. Conversely, if you want to make sure ChatGPT doesn’t remember something, you can use the Temporary Chat feature. On top of that, you can archive chats or full-on delete chats, which removes that conversation from OpenAI servers after 30 days. (It’s worth noting that OpenAI can share your personal info with law enforcement if legally requested.)

All that applies to the information that OpenAI knows about your ChatGPT prompts. What about what others may know about your searches? If your job offers ChatGPT, can your boss see your ChatGPT queries? What about your coworkers? Administrators? Here’s what we know.

How About At Work? Can My Boss See My ChatGPT Queries?

For those wondering if their boss can read their ChatGPT queries, the answer is mixed. Your boss can’t pull up your ChatGPT logs, but the app isn’t exactly a steel trap.

We spoke with system administrators who have experience with ChatGPT’s enterprise product and we learned that managers are not able to view other people’s chats unless those chats are intentionally shared by the user (you can do this by sharing a link using the share icon in the top right or just screenshotting your page). For example, an admin can’t successfully ask ChatGPT, “show me everything Xavier has searched for.”

However, here’s where things get murky. With the right prompts, an admin of your ChatGPT workspace could get vague information about the things other colleagues have been searching for. That could look like, “how many times has Bob asked about Tracy,” and ChatGPT could respond saying, “Bob has asked about Tracy 10 times.” Very suspicious, Bob.

Admins may not have access to worker ChatGPT queries directly but don’t expect the AI tool to successfully keep any info you feed it a secret. ChatGPT has a weird knack for revealing information it learned elsewhere. Beyond the Amazon and Samsung examples above, ChatGPT revealed a journalist’s email address to a PhD candidate at Indiana University Bloomington. The journalist’s email wasn’t a secret, but it revealed the kind of privacy precautions AI tools can lack. As the article points out, ChatGPT isn’t searching the web for its answers, but instead searching all it has learned — sometimes including what it has discussed with others. With the right question, ChatGPT can unearth information others assumed was private, forgotten, or both.

Along with asking the right questions, it may be possible to use a third-party service to check worker activity on ChatGPT. These tools aren’t built into ChatGPT, but learning about other workers’ prompts via a third-party tool is certainly doable.

What Do OpenAI’s Official Docs Have To Say About This?

Looking at official ChatGPT documents, OpenAI’s frequently asked questions section explicitly addresses who can see user chats in the “who can view conversations and chat history” section of the company’s enterprise FAQ. Here’s what the company currently says:

“Within your organization, end users can view their own conversations. Your organization has control over workspaces, and workspace admins can access an audit log of conversations and GPTs through the Enterprise Compliance API. Authorized OpenAI employees will only ever access your conversations for the purposes of resolving incidents, recovering end user conversations with your explicit permission, or where required by applicable law.”

Taking a trip to OpenAI’s Enterprise Compliance page reveals that permissions allowing visibility of a worker's chat logs are typically granted only during “litigation, investigations, or audits.” That means you’re safe to trash talk your boss and employer, and fill ChatGPT’s text box with all sorts of private information, right? Well, not quite. Keep in mind that others can still find out information about your searches if they ask ChatGPT the right questions. Also know that some OpenAI notes in its “Who can view conversations and chat history” section that employees can see your data, “for the purposes of resolving incidents, recovering end user conversations with your explicit permission, or where required by applicable law.”

What should you do to stay private? Here are some tips.

There could be good reasons to go incognito with your ChatGPT usage. Maybe you genuinely don’t want ChatGPT to hold onto a piece of information, or you’re using a friend’s account and you don’t want it to affect their future ChatGPT experience. Whatever the reason, here are a few tips on how to use ChatGPT more privately.

Your first option is the Temporary Chat feature. When Temporary Chat is switched on, OpenAI says anything you type won’t be added to memory, won’t be used to train the AI model and mostly won’t live in your history (OpenAI potentially holds onto the chat for 30 days for “safety purposes”). Another option is changing your settings so that your ChatGPT queries aren’t used to train OpenAI’s models or log any history. You can do that by going to Settings then Show Data Controls and switching Chat History & Training to the off position. Both, presumably, would help users get around the whole “others can learn what I’ve been talking about by asking the right questions” thing.

These AI tools are new and we think companies working on generative AI should prioritize privacy. The idea that, with the right prompt, a person could unearth information from others’ chats is unnerving. Although, when we asked ChatGPT its thoughts, it promised to keep our chats a secret. It’s a good thing generative AI never lies.

Can My Boss Read My ChatGPT Queries?

Written By: Xavier Harding

Edited By: Audrey Hingle, Lindsay Dearlove, Tracy Kariuki

Art By: Shannon Zepeda


Relatearre ynhâld