Worried about privacy when using public AI models? Here are several ways how to keep your conversations with public AI chatbots private.

As discussed by Lance Whitney on ZDNet (How to keep your conversations with ChatGPT, Gemini, Copilot or Claude as private as possible, available here), AI companies don’t always make it easy, but they do offer controls through which you’re able to manage your own privacy.

Whitney discusses how to make conversations with ChatGPT, Google Gemini, Microsoft Copilot and Claude AI more private, with five ways for each platform. The usefulness of each method may depend on what you’re trying to accomplish; for example, the first option for ChatGPT is “Use ChatGPT without an account” – if you already have one, you can log out beforehand. As he notes: “Certain advanced features won’t be available, and you’ll hit lower thresholds.” Still, that could be a useful method in certain circumstances.

Each AI model provides an option to disable the ability for the AI model to train on your conversations. By default, all models train on your content.

Note: Whitney and others have said that Claude doesn’t train on your data by default (and it appears that the default used to be to not train on your data as was the case when I wrote about it here). To determine whether or not Claude defaults to training on your data, I created a new account as I was writing this post. During setup, Claude gives you a couple of options, including whether you want the model to train on your data. It is set to on, meaning unless you change it, Claude will train on your data.

So, I stand by my statement that all models train on your content by default. It’s probably safest to assume that they do anyway.

If you don’t want to turn off training most of the time, but want certain conversations to be private, most of the models provide a temporary or incognito mode to keep the model from training on a sensitive chat. You can also delete a conversation in any model (that doesn’t delete the training of the model if training was enabled).

One area to be very wary of is sharing conversations. As Whitney points out (and I covered on last week’s Kitchen Sink), Claude’s “Share” feature, which generates public links, was enabling the shared content to be indexed by search engines and sensitive content was showing up in Google search results. Whitney says that “Anthropic quickly fixed the ‘exploit’ to disable this type of indexing.” Of course, that doesn’t help those who have already had their information exposed.

Also, given Copilot’s ability to connect to external services such as Microsoft OneDrive, Outlook, Google Drive, Gmail, and Google Calendar, that’s an area unique to it in terms of privacy – you can turn off the switch to any service if you’re concerned about access.

While these AI models are always changing – which means the steps to keeping conversations private will also change over time – this article is a terrific resource for how to keep your conversations with public AI chatbots private!

So, what do you think? Are your conversations with public AI models being used to train the model? Are you sure? 😉 Please share any comments you might have or if you’d like to know more about a particular topic.

Image created using DALL-E 3, using the term “robot using a workstation with tape over its mouth”.

Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.