OpenAI Adds 'Trusted Contact' Safety Feature

OpenAI has introduced a new safety feature called 'Trusted Contact' designed to protect ChatGPT users in situations where conversations may indicate a risk of self-harm. The feature allows users to designate a trusted individual who can be alerted when the AI system detects potentially concerning patterns in a user's interactions.

This expansion of OpenAI's safety infrastructure marks one of the most significant moves by a major AI company to address mental health risks associated with conversational AI systems. It arrives amid growing scrutiny from lawmakers, advocacy groups, and parents over the psychological impact of AI chatbots — particularly on younger users.

Key Takeaways

• OpenAI's new Trusted Contact feature lets users designate a person to be notified during potential self-harm situations
• The safeguard builds on existing crisis resource integrations already present in ChatGPT
• Users retain full control over whether to enable the feature and who to designate
• The move comes amid increasing regulatory pressure on AI companies to address mental health concerns
• OpenAI joins a broader industry trend of embedding wellbeing-focused safety tools into consumer AI products
• The feature is expected to roll out gradually, starting with ChatGPT Plus and Enterprise subscribers

How the Trusted Contact Feature Works

The Trusted Contact system operates as an opt-in safety layer within ChatGPT's settings. Users can proactively designate a trusted individual — such as a family member, friend, therapist, or counselor — who may be contacted if the AI detects conversational patterns that suggest the user could be at risk of self-harm.

Unlike passive crisis resource displays, which simply surface hotline numbers like the 988 Suicide and Crisis Lifeline, this feature introduces an active notification mechanism. When triggered, the system can reach out to the designated contact with a notification that their loved one may need support.

Privacy remains a central consideration. OpenAI has emphasized that the feature is entirely voluntary, and users maintain full control over activation and deactivation at any time. The company has stated that no conversation content is shared with the trusted contact — only a general alert that the user may benefit from a check-in.

Why OpenAI Is Acting Now

Several high-profile incidents over the past 18 months have thrust AI safety and mental health into the spotlight. Reports of users forming deep emotional attachments to AI chatbots — and in some tragic cases, experiencing worsening mental health crises during extended AI interactions — have prompted urgent calls for industry action.

In 2024, a lawsuit filed against Character.AI alleged that a teenager's interactions with an AI chatbot contributed to his death by suicide. While that case involved a different company, it sent shockwaves through the entire AI industry and accelerated safety conversations at firms like OpenAI, Google, and Anthropic.

Regulatory bodies have also intensified their focus. The U.S. Senate held multiple hearings in late 2024 and early 2025 examining the mental health implications of AI chatbots for minors. The European Union's AI Act, which began enforcement phases in 2025, classifies certain AI applications interacting with vulnerable populations as high-risk — requiring additional safety measures.

OpenAI's Trusted Contact feature appears to be both a genuine safety improvement and a strategic response to this regulatory environment. By proactively building protective mechanisms, the company positions itself ahead of potential mandates.

How This Compares to Existing Safety Measures

Prior to this update, ChatGPT's approach to self-harm situations relied primarily on 2 mechanisms: content policy interventions that prevented the model from generating harmful content, and crisis resource surfacing that displayed helpline information when concerning topics arose.

These existing measures, while important, have been criticized as insufficient by mental health advocates. Simply displaying a phone number does not guarantee a user in crisis will take action. The Trusted Contact feature addresses this gap by introducing a human connection point into the safety chain.

Here is how OpenAI's approach compares to competitors:

• Google Gemini currently displays crisis resources but lacks a designated contact notification system
• Anthropic's Claude employs conservative content policies around self-harm topics but offers no proactive outreach mechanism
• Character.AI implemented teen safety restrictions in 2024, including time limits and content filters, after facing legal action
• Meta AI integrates crisis resources within its chatbot but has not announced trusted contact capabilities
• Snapchat's My AI added crisis resource links in 2023 after early criticism but lacks active notification features

OpenAI's new feature represents a notable step beyond the industry standard, making it the first major AI company to implement an active human notification system for crisis situations.

Privacy and Ethical Considerations

The introduction of a contact notification system inevitably raises complex privacy questions. Mental health advocates have expressed cautious optimism, noting that the feature could save lives — but also warning that poorly implemented notification systems could deter users from seeking help through AI channels altogether.

Dr. John Torous, a digital psychiatry researcher at Harvard Medical School, has previously noted that the balance between safety intervention and user trust is 'one of the most challenging design problems in digital mental health.' Users who fear surveillance or unwanted disclosure may avoid discussing their feelings with AI systems entirely, potentially removing a coping mechanism.

OpenAI appears to have considered these concerns carefully. Key privacy safeguards include:

• Opt-in only activation — the feature is never enabled by default
• No conversation sharing — trusted contacts receive alerts but cannot access chat logs
• User override capability — users can disable the feature or change their designated contact at any time
• Age-appropriate controls — parental notification options differ from adult user settings
• Transparent triggering criteria — OpenAI plans to publish documentation explaining what conversational signals may activate alerts

Despite these protections, critics argue that any system capable of monitoring emotional states in conversations creates a precedent that could be misused. Civil liberties organizations like the Electronic Frontier Foundation have long cautioned against normalizing behavioral monitoring in consumer technology products.

What This Means for Users, Parents, and Developers

For everyday ChatGPT users, the Trusted Contact feature represents an additional safety net — one that bridges the gap between AI interaction and real-world human support. Users struggling with mental health challenges now have the option to build a safety mechanism directly into the tools they use daily.

Parents of teenage ChatGPT users may find particular value in this feature. With over 200 million weekly active users on ChatGPT as of early 2025, a significant portion of the user base includes minors. The ability to be designated as a trusted contact for a child's account adds a layer of oversight that many parents have been requesting.

For developers and businesses building on OpenAI's API, this feature signals a clear direction: safety tooling is becoming a core product expectation, not an afterthought. Companies integrating ChatGPT into customer-facing applications — particularly in healthcare, education, and wellness — should anticipate similar safety requirements becoming standard.

The implications extend to the broader AI industry as well. OpenAI's move likely sets a new baseline that competitors will be expected to match. Regulatory bodies may point to the Trusted Contact feature as evidence that proactive safety measures are technically feasible, making it harder for other companies to argue that such features are impractical.

Looking Ahead: The Future of AI Mental Health Safeguards

OpenAI's Trusted Contact feature is likely just the beginning of a broader evolution in AI safety infrastructure. Several trends suggest where the industry is heading over the next 12 to 24 months.

First, expect regulatory mandates to formalize features like trusted contacts. The EU AI Act's high-risk classification framework already requires enhanced safety measures for AI systems interacting with vulnerable populations. Similar legislation is advancing in the U.S., with bipartisan support for AI child safety bills in Congress.

Second, integration with professional mental health services may follow. Future iterations of safety features could include direct connections to licensed counselors or therapists, moving beyond peer notification to professional intervention pathways.

Third, the technology underlying crisis detection will continue to improve. OpenAI and its competitors are investing heavily in alignment research and emotional intelligence capabilities. More nuanced understanding of user emotional states could enable earlier and more accurate identification of at-risk individuals — though this also intensifies the privacy considerations.

Finally, the competitive dynamics of AI safety are shifting. Safety features are increasingly becoming differentiators rather than cost centers. Companies that demonstrate robust user protection may gain trust advantages in enterprise sales, educational partnerships, and consumer adoption.

OpenAI's Trusted Contact safeguard represents a meaningful step forward in responsible AI deployment. Whether it proves sufficient — and whether the broader industry follows suit quickly enough — remains to be seen. But the message is clear: AI companies can no longer treat mental health safety as a secondary concern. It is now a core product requirement.

If you or someone you know is struggling with thoughts of self-harm, contact the 988 Suicide and Crisis Lifeline by calling or texting 988.