Introducing Trusted Contact in ChatGPT

The announcement is quiet on what actually constitutes a triggering signal, who audits that threshold, and whether OpenAI retains any liability if the alert fails to reach a contact in time. The feature also requires users to opt in and designate someone, meaning the population most at risk may be least likely to configure it.

This sits uncomfortably alongside OpenAI's behavioral tracking rollout covered here from The Decoder on May 2nd, where the company enabled ad-targeting data collection by default for free-tier users. That story established a two-tier model where free users carry more exposure. Trusted Contact, which also requires user action to activate, follows the same structural logic: safety infrastructure is available but not guaranteed, and the burden falls on the user. Together, these moves suggest OpenAI is building consumer-platform habits, both monetization and social-safety features, onto what began as a research product, without fully resolving the consent and reliability questions either approach raises.

Watch whether any third-party crisis organization, such as the 988 Lifeline or Crisis Text Line, publicly endorses or audits the triggering criteria within the next six months. Absence of that validation would suggest the feature is more liability management than clinically grounded intervention.