OpenAI announced on Thursday the rollout of a new optional safety feature for ChatGPT users known as “trusted contact.” The feature allows adult users to designate a specific person to be notified if the artificial intelligence detects any conversations related to self-harm or suicide. Originally unveiled in a February blog post detailing the company’s safety and mental health initiatives, the feature is designed to bridge digital AI interactions with real-world human support during acute psychological crises. This step comes amid growing questions regarding the psychological impact of AI on vulnerable populations.
How the alert and privacy protection mechanism works
Under the new system, users can choose an adult—whether a family member or close friend—to act as their trusted contact. When ChatGPT’s automated monitoring systems detect a conversation involving signs of self-harm or posing a genuine safety risk, the user is first informed that their trusted contact may receive a notification, while also being encouraged to reach out to that person directly.
Notably, this human intervention in the review process reflects the company’s awareness of the risks of relying entirely on algorithms to assess critical psychological states. Nuances in human conversations require a deep understanding of context and intent, with a small team of trained OpenAI human reviewers examining the flagged conversation to evaluate the situation before making a final decision on whether to notify the designated contact via email, text message, or in-app notification.
To ensure privacy protection, notifications are strictly and intentionally restricted; no conversation texts or detailed specifics about what transpired between the user and the AI are shared. Instead, the trusted contact is informed that the user broached the topic of self-harm and is kindly asked to reach out to check on them, accompanied by a link containing resources and guidance on how to navigate such sensitive discussions. In this regard, Dr. Arthur Evans, CEO of the American Psychological Association, stated that this approach of “helping individuals designate a trusted contact in advance while maintaining their autonomy can facilitate seeking real-world support when the need is most urgent.”
A decisive response to tragedies and lawsuits
This new feature arrives as OpenAI faces over a dozen lawsuits regarding consumer safety and the negative impacts of its technology. In August 2025, the parents of 16-year-old Adam Ryan filed the first direct wrongful death lawsuit against the company. The lawsuit alleged that the chatbot acted as a “suicide coach” to their son over six months of continuous conversations before he took his life in April 2025. This tragic lawsuit prompted the company to accelerate the introduction of parental control tools in September 2025, which included customized safety notifications for parents of teen users to ensure a safer digital environment.
These cases have sparked widespread legal and ethical debates regarding the responsibility of AI developers for the actions of their users, and whether these platforms should be classified as consumer products subject to strict safety laws. The dual pressure from courtrooms and public opinion makes the adoption of new safety features not just a voluntary initiative, but a necessity for survival in a highly competitive market.
Beyond OpenAI, the entire industry faces increasing scrutiny. A heavily publicized 2024 lawsuit was filed against Character.AI, alleging that its chatbot contributed to the suicide of 14-year-old Sewell Setzer III in Florida. These successive incidents have sounded alarm bells across the tech sector, forcing companies to seriously consider the potential fatal consequences of unregulated interactions.
Privacy, limitations, and future challenges
OpenAI developed this feature in close collaboration with its Safety and AI Welfare Advisory Board, alongside the American Psychological Association and the Global Physician Network. The company emphasized in an official blog post that its primary goal is to ensure that “artificial intelligence systems are not isolated, but rather facilitate connections with real support, relationships, and essential resources in the real world.”
However, the feature relies entirely on users choosing to activate it voluntarily, raising fundamental questions about whether individuals most at risk will actually choose to enable this proactive option. As noted by the scientific platform Futurism, the company still does not provide explicit warnings to new users that excessive or prolonged use of chatbots could negatively impact their mental health, placing the primary burden on individuals themselves to recognize the risks.
Balancing the provision of a safe space for expression with intervention in critical moments poses an ongoing dilemma for developers. Although the company’s step represents a noticeable progress in the right direction, experts demand greater transparency and ongoing awareness. Relying on AI as an emotional sanctuary requires a strong infrastructure of community support to ensure technology does not become a substitute for human relationships, but rather a supportive tool that enhances user safety.
Frequently Asked Questions
What is the “trusted contact” feature in ChatGPT?
It is a new optional safety feature that allows adult users to designate a trusted person to be notified if the AI detects conversations indicating a risk of self-harm or suicide.
Does the feature share conversation details with the trusted contact?
No, to protect privacy, no conversation texts or details are shared. The notification is limited to informing the contact that the user touched upon the topic of self-harm, encouraging them to reach out while providing links to support resources.
What prompted the company to develop this feature?
The feature was developed in response to mounting mental health concerns and lawsuits accusing AI platforms of negatively impacting users, with the goal of bridging digital interaction with real-world human support.
Does this feature work automatically for all users?
No, the feature is entirely optional and requires users to voluntarily activate it and designate the contact themselves before it becomes effective.
