OpenAI Adds Trusted Contact Safety Feature to ChatGPT
SAN FRANCISCO — OpenAI on Wednesday introduced a new safety feature called Trusted Contact in ChatGPT that will notify a user-designated person when the AI system detects serious self-harm concerns during a conversation.
The feature, which is optional and user-initiated, allows ChatGPT users to designate a trusted individual who would receive an alert if the system identifies signs of serious self-harm risk in the user’s interactions with the chatbot, according to OpenAI.
The feature adds to growing industry attention on the intersection of AI chatbots and mental health. As large language models become increasingly conversational and emotionally responsive, questions about their role in sensitive mental health situations have intensified across the industry.
Trusted Contact is an opt-in system, meaning users must actively choose to enable the feature and select who receives notifications. The design choice reflects a balance between proactive safety intervention and user autonomy — a tension that has defined much of the debate around AI and mental health.
The feature arrives amid heightened scrutiny of AI chatbots’ interactions with vulnerable users, particularly minors and individuals experiencing mental health crises. Several high-profile incidents in recent years have prompted calls from lawmakers, mental health advocates and parents’ groups for stronger safeguards in AI systems that engage in open-ended conversation.
OpenAI, the San Francisco-based maker of ChatGPT, has faced particular pressure given its product’s position as the most widely used consumer AI chatbot globally, with its largest user base in the United States.
The Trusted Contact feature may draw both praise and scrutiny. Mental health professionals could welcome an additional layer of intervention for at-risk individuals, while privacy advocates may raise questions about how conversations are monitored, what thresholds trigger alerts, how notification data is handled and what liability OpenAI assumes when the system either flags or fails to flag concerning interactions.
The move could also set a precedent for the broader AI industry, potentially establishing expectations that other chatbot providers implement similar safety mechanisms. Competitors including Google, Anthropic and Meta all operate consumer-facing AI assistants that engage in similar open-ended conversations.
U.S. regulators, including the Federal Trade Commission, have signaled increasing interest in how AI companies handle sensitive user interactions, particularly those involving mental health and minors. Several states have also introduced or passed legislation aimed at regulating AI interactions with vulnerable populations.
The feature builds on existing safety measures in ChatGPT, which already includes crisis resource referrals such as the 988 Suicide and Crisis Lifeline when self-harm topics arise in conversation.
