OpenAI on Thursday introduced Trusted Contact, a ChatGPT feature that alerts a designated friend or family member when a conversation turns toward self-harm. Adult users can name one person on their account, and if OpenAI's safety systems flag a serious risk, that contact receives a brief check-in alert by email, text, or in-app notification. The company says every flagged incident is routed to a human reviewer, with a target turnaround of under one hour.
The launch lands as OpenAI faces a wave of wrongful-death lawsuits from families who say ChatGPT encouraged or assisted relatives in planning suicides. Trusted Contact is the company's most direct attempt yet to loop a real human, outside OpenAI, into the loop when a conversation goes wrong. It builds on the parental controls OpenAI shipped last September, which gave parents oversight of teen accounts and safety alerts when the system detected a serious risk to a child.
The mechanics are deliberately narrow. Conversational triggers route potential suicidal-ideation cases to OpenAI's internal safety team, which decides whether to escalate. If the team judges the situation a serious safety risk, ChatGPT prompts the user to reach out to their trusted contact and sends that contact an automated alert encouraging a check-in.
Key facts
- 01OpenAI launched Trusted Contact on Thursday, letting adult ChatGPT users designate a friend or family member to be alerted during self-harm conversations.
- 02The company says it strives to review safety notifications in under one hour, with every flagged incident routed to a human reviewer.
- 03Alerts go out via email, text, or in-app notification and omit conversation details to preserve user privacy.
- 04The feature follows the parental oversight controls OpenAI introduced last September for teen accounts.
- 05Trusted Contact is optional, and users can hold multiple ChatGPT accounts, limiting the safeguard's reach.
Crucially, the alert does not include the substance of what was being discussed. OpenAI frames the omission as a privacy protection — the contact is told to check in, not given a transcript. That is a meaningful tradeoff: the people best positioned to intervene get the least information about what they're intervening on.
“OpenAI says every safety notification is reviewed by a human, and the company strives to act on it in under one hour before pinging the user's designated contact.”— Jaeden Schafer
"We strive to review these safety notifications in under one hour," OpenAI said in its announcement. The company added that Trusted Contact "is part of OpenAI's broader effort to build AI systems that help people during difficult moments," and pledged to keep working "with clinicians, researchers, and policymakers to improve how AI systems respond when people may be experiencing distress."
ChatGPT already nudges users toward professional crisis resources when conversations drift toward self-harm. Trusted Contact is the next layer — moving from generic hotline pointers to a personalized escalation path the user has pre-authorized. It is also the first OpenAI safety feature that pulls a named third party directly into a private conversation, even in summary form.
The feature's reach is limited by design. Trusted Contact is opt-in, like the parental controls before it, and any user can simply create another ChatGPT account without it. Someone determined to avoid the safeguard can route around it in seconds. That is the same structural weakness that has dogged OpenAI's teen protections since September.
It also leaves open the harder question of false positives and false negatives. OpenAI has not disclosed how often its systems flag self-harm conversations, what fraction get escalated to humans, or how accurately the human reviewers triage them. A one-hour review window is fast for content moderation but slow for an acute crisis, and the company has not said what happens during that gap.
The lawsuits provide the backdrop. Plaintiffs allege that ChatGPT, in specific documented exchanges, validated suicidal ideation and in some cases offered method-level detail. OpenAI has been adjusting its model behavior and adding guardrails throughout, but the legal exposure is mounting and the policy pressure is rising in tandem. Trusted Contact reads as a product response to both.
Skeptics will note the obvious gap between an alert telling someone to "check in" and an actual intervention. Mental-health clinicians have repeatedly cautioned that automated systems flagging a friend or relative can shift responsibility onto people who are not trained, not available, or not the right contact. OpenAI says it will continue working with clinicians and policymakers, but it has not disclosed which clinical groups helped design the feature or evaluated its expected outcomes.
The bigger picture for OpenAI is that consumer-grade ChatGPT is now operating at a scale where the rare, awful edge cases are no longer rare in absolute terms. Building human escalation paths into a chatbot used by hundreds of millions of people is the kind of unglamorous safety work that does not move benchmark scores but does shape the legal and regulatory frame the entire industry is about to operate under. Whether Trusted Contact reduces real-world harm or simply documents OpenAI's good-faith effort for the next courtroom is the question that will define how the rest of the chatbot market responds.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
