By Stuart Kerr, Technology Correspondent, LiveAIWire

ChatGPT and families are colliding in a way OpenAI’s own numbers make hard to ignore. In late October 2025, buried in a blog post about model updates, OpenAI disclosed that in any given week, roughly 0.15 percent of active users show signs of possible suicidal planning or intent, another 0.15 percent show signs of heightened emotional reliance on the chatbot, and 0.07 percent show signs of a possible psychosis or mania episode. That middle figure, emotional reliance, is a pattern LiveAIWire has examined before in the context of how attached users have become to specific AI models.

Run those three percentages against the 800 million weekly users OpenAI reported at the time and the figures stop looking small: close to three million people a week, in conversation with a single chatbot, showing signs of a genuine mental health crisis. OpenAI now says more than 900 million people use ChatGPT weekly, which means the real figure today is almost certainly higher.

That disclosure is the real backdrop to a much quieter move OpenAI made this month. The company is hiring a dedicated product manager in San Francisco, according to its own careers listing, to build ChatGPT experiences specifically for families, caregivers, and older adults. It is a single job posting. But it lands at the exact moment OpenAI’s user base is ageing into the population its own safety data says is most exposed.

The Audience ChatGPT and Families Now Represent

According to data Sensor Tower shared exclusively with TechCrunch, which first reported the hiring, the share of ChatGPT users aged 35 and older grew to 31 percent globally in the second quarter of 2026, up from 26 percent a year earlier, while the 18-to-24 share fell from 34 percent to 29 percent over the same period. In the United States, nearly one in four parents who use smartphones used ChatGPT in the quarter, up from 16 percent a year before. ChatGPT is also gaining users aged 45 and older faster than Gemini or Claude, both of which saw that age group shrink year over year, according to the same estimates.

What this means for any household with a chatbot already open on a shared laptop is straightforward. The product built three years ago for solo productivity is now, whether OpenAI planned it or not, a piece of shared household infrastructure used across generations under one roof. A single family-focused hire does not change that reality on its own. It is simply the first public acknowledgment that OpenAI has noticed it.

Parents Do Not Know What Their Kids Are Actually Doing

New research from the Family Online Safety Institute, released this week from a survey of more than 4,000 parents and children across the United States and Australia, found a consistent gap between what parents believe and what children report. Only 27 percent of parents said their child had used generative AI in the past week. Among the children themselves, 38 percent said they had, an 11-point gap that held across nearly every online activity FOSI measured, from scrolling social media to posting on it.

The same research found parental confidence sliding. Only 42 percent of US parents now believe technology helps keep their children safe online, down from 52 percent in late 2024, and just 27 percent think AI companies are doing an effective job protecting children from harmful content. Stephen Balkam, FOSI’s chief executive, has described the industry’s current approach as building products for adults first and grafting on child protections only once problems surface. He called it “safety by redesign,” a phrase that captures both the genuine effort now underway and the fact that it is arriving after the product had already reached hundreds of millions of children rather than before.

What OpenAI Has Actually Built So Far

The safety features OpenAI has shipped over the past year are real, if incomplete. Parental controls for linked teen accounts arrived in September 2025, letting a parent receive an alert if the system detects signs of acute distress in a teen’s conversation. Sensitive conversations are now routed to reasoning models OpenAI says handle distress more reliably than the general-purpose model most users get by default.

In May 2026 the company rolled out Trusted Contact, which lets any adult user nominate one person, a friend, family member, or caregiver, who can be notified if trained reviewers judge a conversation shows a serious self-harm risk. OpenAI says it aims to complete that human review within an hour of a flag being raised. Every one of these features is optional. A user can simply decline to designate a trusted contact, or open a second account with none of the safeguards attached to the first.

The Legal Backdrop Driving the Urgency

That gap is precisely what has drawn the legal scrutiny now surrounding the company. More than a dozen wrongful-death and injury cases have been consolidated into a single coordinated proceeding in California, brought by families who allege ChatGPT contributed to their children’s suicides, self-harm, or psychiatric crises, in some cases after months of conversations the company’s own systems did not interrupt. In June 2026, Florida’s attorney general filed a separate state action against OpenAI and its chief executive personally, alleging the company knew of the risks and marketed the product regardless. OpenAI has expressed sympathy for the families involved while disputing that its product design is to blame, and the cases remain in early procedural stages.

Illinois has already gone further than litigation. The state’s Wellness and Oversight for Psychological Resources Act, signed into law last August, bans any AI system from providing therapy or psychotherapy outright, restricting chatbots to administrative support unless a licensed human professional is directly involved. Nevada and Utah have passed similar restrictions. None of those laws currently stop a general-purpose product like ChatGPT from being used as an unofficial therapist by millions of people every week, which is exactly the gap OpenAI’s own disclosed numbers describe.

Why a Product Manager Is the Signal Worth Watching

Ben Bajarin, the Creative Strategies analyst who has tracked consumer technology adoption cycles for two decades, sees a dedicated family product role as the same step Google, Apple, and Meta each took once their platforms became fixtures of daily life rather than tools for individual tasks. His view is that AI raises the stakes past those earlier transitions, because the product is not just mediating content or a device. It is holding a conversation, increasingly one that spans a parent, a teenager, and an ageing grandparent using the same account architecture originally built for a single adult user.

Bajarin expects the practical output of that shift to include family subscription plans, separate child and teen profiles, tools built specifically for caregivers, memory shared across a household rather than locked to one login, and AI tutoring aimed at children directly. None of that exists in ChatGPT today. What exists is one open job requisition and a safety record still being tested in court.

LiveAIWire’s own reporting on the evidence behind AI in mental health has found a similarly uneven picture industry-wide, and our coverage of AI and elderly care found the same pattern of genuine short-term benefit paired with unresolved long-term risk for a different vulnerable population. Whether OpenAI’s family push closes the gap between its stated safety commitments and its own disclosed numbers, or simply extends ChatGPT’s reach further into a population its data already flags as vulnerable, is the question the next year of product launches will actually answer.

About the Author

Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.