Built-in AI assistants are popping up everywhere. In your e-mail. In your notetaker. In your CRM. Maybe not in the menu in your favorite restaurant yet, but... give it time.
You didn't ask for them, but they're genuinely helpful. They revise, or summarize, or organize. And they're free! Whee!
Here's the thing:
When something digital is free, you're usually paying for it with something other than money.
When you use AI, whatever you put in the chat leaves your computer and travels to "the cloud" (the AI's servers, in this case) for processing. And some AIs use your chat contents for "training," the process of improving future AI models. Which means what you paste into your friendly little assistant might not stay just between the two of you.
And THAT should give you pause.
Much of what you share with the AI is innocuous. But if you're a therapist handling confidential patient information, or a business with proprietary IP, you miiiiiight not want those details training the future version of ChatGPT.
(By the way: if you're a therapist who's thinking of pasting PHI into an AI, make sure you have a BAA with the vendor that specifically covers this use. And even if you do... just think carefully about how much you're willing to trust an AI with your patients' data...)
It's not all bad news. You don't have to stop using your AI assistants. But you definitely should figure out whether they're training on your data. And figuring that out is pretty straightforward: a Google search on "does [AI name] in [program] train on my data?" gives you a response like this. Easy peasy.
Also: not all vendors use your data for training. Google and Microsoft specify that their e-mail assistants do not train on your data. Note, however, that's e-mail assistants, not their main AIs. The big consumer chatbots all default to training on your data: Claude, ChatGPT, Gemini, Grok, and Copilot. The versions that promise not to are usually the paid business tiers, like Microsoft 365 Copilot or paid Google Workspace.
So if you don't want training, you may need to switch it off. Search "how do I disable training in [AI name]" for any you use, and Google will show you how to disable training. Adjust the search as necessary if it's an AI assistant in an app. If you care about your data, do this for all your AIs (not just the main one you use).
The bottom line: your information is valuable to these vendors. So your goal for the week: figure out whether you want to limit how they use that information, and then set those limits.
If you got this far, I'd love to hear from you. Which are you limiting first?
