The company behind ChatGPT is simultaneously rolling out new emotional-safety features and navigating a growing wave of litigation. Here’s what both developments tell us about where AI companionship is headed.
With over 900 million weekly users, ChatGPT has become one of the most significant mental health-adjacent products ever deployed, not by design, but by sheer scale. People confide in it, lean on it during difficult moments, and increasingly treat it as a first port of call for emotional support. That reality is now forcing OpenAI to reckon with two simultaneous pressures: building better safeguards, and defending itself in court over the ones it didn’t have earlier.
New Protections, Carefully Worded
OpenAI recently published an update on its mental health-related safety work, and the announcements are incremental, but meaningful.
The company has been building on parental controls introduced in September 2025, which have seen what it describes as “encouraging engagement from families.” More notably, OpenAI is preparing to launch a trusted contact feature for adult users — a mechanism that would allow someone to designate a person to receive notifications if the system detects they may need additional support. Think of it as a digital safety net: an opt-in layer that connects the platform to real-world human relationships when the AI flags potential distress.
This feature is being developed in collaboration with OpenAI’s Council on Well-Being and AI and its Global Physicians Network — a signal that the company is treating this as a clinical challenge, not just a product problem.
On the technical side, OpenAI says it is expanding its evaluation methods to include simulations of extended mental health-related conversations. The goal is to better detect where ChatGPT’s responses might fall short — or even cause harm — in sensitive, sustained interactions. Details on the specific methodology are still pending, but the company says more will follow in coming weeks.
The Legal Reckoning
The timing of this update is not coincidental. Running alongside the safety announcements is a significant litigation development: a California court has consolidated multiple mental health-related cases involving ChatGPT into a single coordinated proceeding. New cases are expected to be filed as part of that process.
OpenAI’s public response strikes a careful, empathetic tone, outlining principles of good-faith engagement with the facts, sensitivity toward individuals involved, and a commitment to understanding complex circumstances before drawing conclusions. It explicitly acknowledges that these are “heartbreaking situations” and urges patience as the court process unfolds.
What the company is navigating is a genuinely difficult question the entire AI industry will eventually face: when an AI system is embedded in someone’s emotional life, and something goes wrong, where does responsibility lie?
Legal proceedings rarely produce clean answers, and in this domain — where causation is entangled with pre-existing conditions, platform design decisions, and individual vulnerability — they may produce even fewer than usual. But the consolidation of cases into a single California proceeding suggests the judicial system is beginning to treat this as a systemic issue, not a collection of isolated incidents.
What This Means for the Industry
OpenAI is not alone in facing this tension. Any AI platform that allows unstructured, open-ended conversation at scale will eventually encounter users in crisis. The question of how to respond, technically, ethically, and legally is one the industry has largely deferred.
The trusted contact feature, if it works as described, represents a meaningful shift in thinking: rather than treating the AI as a self-contained safety system, it tries to connect distressed users back to human support networks. That’s a philosophically important distinction, and one worth watching as implementation details emerge.
The litigation, meanwhile, will likely define the contours of AI platform liability for years to come, not just for OpenAI, but for every company building systems that people talk to when they’re struggling.
This article is based on a public update published by OpenAI regarding its mental health-related safety features and ongoing litigation.
Read the Open AI blog here: https://openai.com/index/update-on-mental-health-related-work/
