Have you ever been using ChatGPT to edit a piece of content or create an image, only to realise you’ve just shared more information than you should have? This might be your age, a real photo of a loved one, or perhaps even the name of your company. 

While it may seem harmless at first, your data could come back to bite you in the algorithm later down the line.

Where does your conversation data go?

The bad news is, whatever you’ve told ChatGPT exists forever… and it isn’t private. At the moment, it’s kept on file by OpenAI and used for a variety of purposes, including quality improvements, iterative learning, and base model training.

If you asked ChatGPT to proofread a short, non-sensitive email, no big deal. But if you’ve spoken about a recent break-up, shared a CV with all your contact information on, or asked for advice about a personal medical issue, these conversations won’t disappear. In fact, 11% of all data pasted by employees into ChatGPT in 2023 was found to be strictly confidential.

And while your conversations aren’t being officially sold on to advertisers (or worse) yet, that’s not to say they couldn’t be in the future.

Is your data being sold?

Sorry, did we say that your data isn’t being sold to advertisers? Well, officially it’s not – but some users have still noticed some troubling patterns.

“I recently had a conversation with ChatGPT about what would be involved in creating an app. Before or after this conversation, I did not look up any more information about this… shortly thereafter I was served an ad on YouTube by sentry.io for their Mobile APM.” One user recounted on Reddit.

Another said: “Recently I was confused about my situation in life. I asked ChatGPT if it was ADHD, my first encounter of talking about it digitally… [afterwards] I started getting ads about ADHD quizzes, and products, and cognitive behavioural therapy and things.”

This isn’t down to an error in the system – it’s a sneaky loophole that allows OpenAI to disclose information to law enforcement, government agencies, and affiliates. Who are those affiliates, you ask? That’s for OpenAI to know and you to find out when they serve you a highly targeted ad.

The dangers of oversharing to ChatGPT

But oversharing couldn’t just land you in hot water when it comes to your cyber health – it also poses a serious risk to your mental wellbeing. Physicians and psychologists have observed an emerging condition they’re calling “AI psychosis”, in which patients treat AI as a romantic partner, religious figure, or even a cult-like follower.

Unable to engage in a serious, two-sided conversation, the chatbots are only capable of reinforcing the user’s existing viewpoints and amplifying their delusions. In one of the most infamous online cases, a woman named Kendra recounted how she fell in love with her human psychiatrist, then turned to two AI bots to act as her “virtual psychiatrists” who now help share her message on social media.

But not every case of AI psychosis is quite as TikTok-worthy. A man from Scotland asked ChatGPT for legal advice after he was let go from work – as he shared more information, the chatbot began to tell him that he was in line for a big payout and a movie deal based on his experience, and that he would imminently become a multi-millionaire as long as he followed its (completely untrained) advice. Eventually, he had a breakdown and only regained his grip on reality when he saw a real therapist.

Can it be used ethically? Maybe

This doesn’t change the fact that, for millions of people, AI is a lifeline. “For me, as an older person, it’s helping to keep my brain active,” says Fasthosts Partner and serial founder Keith Grinsted. “I chat to it in order to ward off dementia.”

Others have found it to be a more efficient translation assistant than something like Google Translate, while frequent gym-goers use it to prepare meals and plan their workouts without having to spend hundreds of pounds on a personal trainer.

But the fact is that there’s no reliable, or truly ethical, replacement for real human connection. That’s something we advocate for on both a personal and professional level.

Fasthosts: Human-led, always

When you decide to host your website or source your domain through Fasthosts, you can count on award-winning expert support around the clock. Our human agents are available 24/7 to answer your calls, emails, or social media DMs – so you never have to turn to ChatGPT for answers.

Plus, we go even further for our Partners, assigning a dedicated Partner Manager to each account to support your referrals, rewards, and queries. Get in touch with our team to find out how else we put humans above bots at Fasthosts.