OpenAI Rolls Out ChatGPT Trusted Contact Feature to Bridge AI and Real‑World Support

OpenAI announced on Thursday a new safety layer called ChatGPT Trusted Contact, letting users name an adult who receives a discreet alert if the model detects language suggesting self‑harm.

How the feature works

Within the settings menu, a user adds a contact over 18 years old via a short form; the nominee need not own a ChatGPT account. OpenAI then sends an invitation that must be accepted within ↑ 1 week. If the invite is declined, the user can nominate someone else.

Trigger and response

When the AI flags a conversation as potentially suicidal, an internal review team—described as “small and specially trained”—confirms the risk before a brief email or text reaches the trusted contact. The message contains no transcript, only a note that the user mentioned self‑harm and a link to best‑practice guidance.

“Helping people identify a trusted person in advance, while preserving autonomy, can make real‑world outreach possible when it matters most,” said Dr. Arthur Evans, APA chief executive.

The rollout follows a series of lawsuits alleging OpenAI’s model failed to intervene in fatal cases. OpenAI denies liability, but the company says the new tool aligns with its goal that “AI systems should not exist in isolation.”

Experts from the American Psychological Association and OpenAI’s own Expert Council on Well‑Being consulted on the design. Florida’s attorney general is reportedly reviewing whether the feature satisfies state‑level obligations on suicide prevention, as reported by Reuters and AP News.

Users may edit or delete the contact at any time; contacts can also remove themselves via the help center. OpenAI emphasizes that the service is optional and that immediate crisis resources—such as the 988 Suicide & Crisis Lifeline—remain the first line of defense.

Intel provided by Dr. Aris Thorne (Artificial Intelligence Researcher).