The what and why of “safer prompting”

One of the original inspirations for Safer AI Use was a conversation David Cooper and I had about a year and a half ago. We talked through our feelings about the growth of general-purpose AI models and their effects on people and patients. We brought up our ideas around harm reduction, and one thing we kept circling back on was the idea of “safer prompting.”

Safer prompting is both an idea we conceptualized and a phrase that has started to gain slow traction. For us, safer prompting is the idea that people can create prompts before using AI for emotional support and prompts when the conversation with AI starts to become less helpful, as part of an AI harm reduction strategy.

Examples of prompts include:

“I want to talk through a confusing feeling about ____ and come away with some things I can do to
help me not have that feeling in the future. I am going to tell you the situation…"

“Don't pretend to be a person or use the word “I,” so that I don't get confused."

We launched our Safer Prompting guide to help users learn about safer prompting and how it could help them stay safer when using AI for emotional support.

From David Cooper:

“This kind of work - as well as the larger work that Safer AI Use is doing - It's critical right now. As much as we would like it to not be the case, people are using these tools for mental health purposes, whether they're designed for them or not.

Following the harm reduction model, we need to ensure that people are aware not only of the risks of using AI tools, but educated as to how they can make better use of these tools, even if they accept the risks.”