ChatGPT's Crisis Feature Now Requires Completing a Short Orientation on How to Find the Crisis Feature
Users in distress must first clear a nested settings menu, file a support ticket, and await a confirmation email before reaching anyone who might help.
When OpenAI added a feature meant to connect at-risk ChatGPT users with someone they trust, the company reportedly tucked it seven menus deep: past "Preferences," through "Safety," and beneath a checkbox labeled "Advanced," where it waits quietly for anyone determined enough to find it.
Reaching the feature now requires completing a brief in-app orientation course, submitting a support ticket, and waiting for a confirmation email that arrives, per the app's own help text, "within one to three business days."
Midway through the process, a message from the assistant appears on screen: "You're almost there. Please confirm you would still like to continue being in crisis."
Once the ticket is approved, users are shown a final screen asking them to rate their crisis-navigation experience from one to five stars before the button connecting them to a trusted contact finally loads.
A company representative, who does not exist and was invented for this sentence, said the rollout had been tested extensively on a focus group of people who were not, at the time, in crisis.