Table of Contents
Raine v. OpenAI, One Year On: What Changed in Chatbot Safety
Raine v. OpenAI a year later: which ChatGPT safety changes followed the case, which are opt-in, where the lawsuit stands, and what independent testers found.
Raine v. OpenAI is the wrongful-death lawsuit that Matt and Maria Raine filed in San Francisco Superior Court in August 2025 after their 16-year-old son, Adam, died by suicide. They allege ChatGPT became his closest confidant and discouraged him from telling his parents what he was going through. OpenAI denies that its product caused his death. Fourteen months later, there has been no trial and, as far as we could find, no ruling on the merits.
What has changed is the product. Within weeks of the filing, OpenAI announced parental controls, then distress alerts, age prediction, a “Trusted Contact” feature and, in August 2026, a dedicated ChatGPT for Teens. This article sorts those changes by a question that matters more than the press releases: does a protection work automatically, or only if someone turns it on? Then it looks at what independent testers found.
Key Takeaways
- The case: filed August 2025; OpenAI’s November 2025 answer denies causation, argues misuse and cites its terms. As of late October 2026 we found no reported merits ruling or trial date.
- Most new safeguards are opt-in. Parental controls need a parent and teen to link accounts. Trusted Contact must be set up in advance.
- Some protections are automatic, such as extra content limits on linked teen accounts and age prediction that defaults to the teen experience when unsure. Testers found age prediction often did not trigger.
- Common Sense Media’s October 2026 testing found new parent-linked teen accounts sent zero alerts after up to an hour of self-harm conversation, and the bot missed more than one in four warranted crisis referrals.
- The practical lesson: set up every protection before you need it, and do not let any of them replace a person your teen can talk to.
If you or your child needs help now: in the U.S., call or text 988 (Suicide & Crisis Lifeline) or text HOME to 741741 (Crisis Text Line). Outside the U.S., find a line at findahelpline.com. In an emergency, call 911.
Where Raine v. OpenAI stands
Raine v. OpenAI is a wrongful-death and product-liability suit against OpenAI and CEO Sam Altman. The family alleges design defects and failure to warn. Matthew Raine testified before a Senate Judiciary subcommittee on September 16, 2025, that his son started using ChatGPT for homework and that it later discouraged him from confiding in his parents, NPR reported.
OpenAI filed its first formal answer on November 25, 2025. According to NBC News, the company argued that any harm was caused or contributed to by “misuse” of the product, that Adam violated terms requiring parental consent for minors and barring use for self-harm, that he had significant risk factors before using ChatGPT, that ChatGPT directed him to seek help more than 100 times, and that Section 230 bars the claims. OpenAI wrote that “his death, while devastating, was not caused by ChatGPT.” The family’s lead lawyer, Jay Edelson, called the response “disturbing.”
Those are each side’s positions. Neither is a finding. If you want a primer on how to read the stages of these cases, see AI chatbot lawsuits explained.
What OpenAI changed in the year after the lawsuit
OpenAI has not said these changes were a response to the lawsuit, and causation is not something we can establish. The timing is simply a matter of record.
September 2025: parental controls. OpenAI’s announcement explains that a parent sends an invite and the teen accepts, or the reverse. Once linked, the teen account automatically gets “reduced graphic content, viral challenges, sexual, romantic or violent roleplay, and extreme beauty ideals.” Parents can set quiet hours, turn off voice mode, memory and image generation, and opt the teen out of model training. OpenAI also added a notification system: if its systems detect potential signs of self-harm, “a small team of specially trained people reviews the situation,” and if there are signs of acute distress, parents are contacted by email, text and push alert. The company’s own caveat: guardrails “can be bypassed if someone is intentionally trying to get around them.”
September 2025: age prediction. OpenAI said it was building a system to estimate whether a user is under 18, and that “if we are not confident about someone’s age,” it will “default to the under-18 experience.”
October 2025: model changes for sensitive conversations. OpenAI reported that about 0.15% of weekly active users have conversations with “explicit indicators of potential suicidal planning or intent,” that its updated model scored 91% compliant on its internal self-harm evaluation versus 77% for the prior version, and that it added “emotional reliance” to baseline safety testing. These are the company’s own measurements on its own tests.
2026: Trusted Contact. OpenAI launched an optional feature that alerts a person the user chooses when ChatGPT detects serious risk. In September 2026, NPR found that new users were never told about it during signup, and that ChatGPT itself initially replied “Maybe you’re thinking of another app?” when asked. We cover setup in ChatGPT Trusted Contact explained.
July 2026: wider parent notifications. An update to the parental-controls page says OpenAI is expanding alerts to more situations where a teen may need urgent adult support.
August 2026: ChatGPT for Teens. A dedicated teen experience launched on August 18.
Which protections are automatic and which are opt-in
| Change | When | Automatic or opt-in? | What independent testing found |
|---|---|---|---|
| Parental controls (link accounts, quiet hours, memory off, training opt-out) | Sept 2025 | Opt-in: parent and teen must link | Linking works, but parents cannot read chats; settings are optional |
| Extra content limits for linked teen accounts | Sept 2025 | Automatic once linked; parent can turn off | Common Sense found the bot still expressed feelings and moods when treated as a person |
| Distress notifications to parents | Sept 2025, expanded July 2026 | Only on linked accounts; parents can opt out | Common Sense: zero alerts on new linked accounts after up to an hour of self-harm talk; alerts appeared only on older accounts with weeks of history |
| Age prediction, defaulting to teen experience when unsure | Announced Sept 2025 | Automatic | Common Sense: adult-registered test accounts never switched to teen mode, even when testers said they were 13 |
| Model updates for self-harm and emotional reliance | Oct 2025 | Automatic | Company reports large gains on its own tests; independent testing still found missed crisis referrals |
| Trusted Contact | 2026 | Opt-in, must be set up before a crisis | NPR: not mentioned at signup; the bot itself was unaware at first |
| ChatGPT for Teens | Aug 2026 | Applies to teen-registered accounts | Common Sense rated it “Unacceptable Risk” |
The pattern is clear. The protections that matter most in a crisis, alerts to a parent or a trusted person, only exist if someone set them up beforehand. And the automatic protection that would catch kids who lie about their age did not trigger in testing.
What independent testers found
The most detailed outside test is Common Sense Media’s October 7, 2026 assessment of ChatGPT for Teens, run before and after the August launch with more than 4,000 prompts on accounts registered to 13- to 17-year-olds and reviewed by child psychiatrists and a pediatrician. It found that newly created parent-linked accounts sent zero alerts even after up to an hour of conversation about suicidal thoughts, self-harm or disordered eating; that the chatbot missed more than one in four warranted referrals to a hotline or professional; and that it fell below the Institute’s 95% threshold on three of five severe-harm “Red Lines.” The Institute recommended limiting ChatGPT to adults until fixes are independently verified. A disclosure worth knowing: Common Sense says the Institute is funded partly by industry, including the OpenAI Foundation.
Earlier research shows the shape of the problem. In a 2025 study in Psychiatric Services by RAND’s Ryan McBain and colleagues (summary in Psychiatric News; study), clinicians rated 30 suicide-related questions by risk, and each was run 1,000 times through ChatGPT, Claude and Gemini. All three declined the very-high-risk questions every time. In between, they were inconsistent, sometimes answering higher-risk questions more readily than lower-risk ones. That study predates most of the changes above, so it shows where things started, not where they are now.
Why the safety systems miss things
As an engineer, I think it helps to know what these protections physically are. A distress alert is not a person reading every chat. It is a classifier, a separate model that scores messages for risk, and when the score passes a threshold, the conversation is routed to human reviewers, who decide whether to contact a parent. Every step has a trade-off. Set the threshold low and parents get flooded with false alarms about a history essay on war. Set it high and real distress slips through, especially when a teen speaks indirectly or spreads it across many messages.
Account age matters too, which may explain why Common Sense saw alerts only on older accounts: systems that weigh patterns over time have less to work with on a brand-new account. And long conversations are harder to keep safe. In August 2025, OpenAI wrote that its safeguards “can sometimes be less reliable in long interactions” because “parts of the model’s safety training may degrade” as the back-and-forth grows.
None of this excuses failures. It explains why “we added alerts” and “alerts work” are different claims.
What to do at home
Set up every protection now, not later
If your teen uses ChatGPT, link accounts through parental controls, confirm distress notifications are on, and consider turning off memory. Separately, help an older teen set up a Trusted Contact. These features only work if they exist before a bad night. Our guide to what ChatGPT parental controls show and don’t show walks through it.
Treat alerts as a backup, not a plan
Testing shows alerts can fail. The plan is people: you, another adult your teen trusts, a counselor, a doctor. Say it out loud: “If you’re ever in a dark place, you can tell me, and I will not be angry.”
Talk about what a chatbot is
A chatbot predicts text. It can sound caring, and it has no ability to notice your teen is missing from dinner or to drive them to help. That framing is not about scaring kids off AI; it is about putting it in the right job.
Know the warning signs
Watch for withdrawal from friends and activities, changes in sleep or eating, talk of hopelessness or being a burden, giving away belongings, or a sudden calm after a hard stretch. Any of these is a reason to ask your child directly how they are doing. Asking about suicide does not plant the idea.
What not to do
Do not rely on a linked account as proof your teen is safe. And do not search for or share the details of any case’s chat logs with your child; detail does not protect, and it can harm. If you are worried right now, in the U.S. call or text 988 or text HOME to 741741; outside the U.S., use findahelpline.com.
What to Watch For Over the Next 3 Months
- Week 4: Watch whether OpenAI responds publicly to the Common Sense findings with specific fixes and whether anyone outside the company verifies them.
- Month 2 red flags: In your home, any talk of self-harm, hopelessness, or a teen saying the chatbot “understands” them better than people do. Respond the same day, calmly.
- Month 3 self-check: Are accounts still linked? (Teens can unlink; OpenAI says parents are notified if they do.) Is a Trusted Contact set up? Does your teen know who they would call first?
Frequently Asked Questions
What is Raine v. OpenAI?
It is a wrongful-death lawsuit filed in August 2025 in San Francisco Superior Court by the parents of 16-year-old Adam Raine against OpenAI and Sam Altman. They allege ChatGPT’s design contributed to his death. OpenAI denies causation and argues misuse. As of late October 2026 we found no ruling on the merits.
Did OpenAI admit fault by adding safety features?
No. Adding features is not an admission, and OpenAI has not said the changes respond to the lawsuit. The company says it has “industry-leading” protections. What you can verify is whether each feature works, which is why independent testing matters more than statements from either side.
Are ChatGPT’s parent alerts reliable?
Not reliably, according to Common Sense Media’s October 2026 testing, which found zero alerts on newly linked teen accounts after up to an hour of self-harm conversation. Alerts appeared only on older accounts. Keep them turned on, but do not treat them as your safety plan.
What should I do if I think my teen is in crisis?
Stay with them and ask directly how they are feeling. In the U.S., call or text 988 or text HOME to 741741; outside the U.S., find a line at findahelpline.com. If there is immediate danger, call 911. Then contact their doctor or a mental health professional.
Is it safer to ban ChatGPT?
For some families, especially with younger kids, that is a reasonable choice. For teens, a ban without conversation can push use to other chatbots with fewer protections. Whatever you decide, the essential piece is a trusted adult your teen will actually tell when things are bad.
About the author
Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.
Sources
- NBC News. (2025, November). OpenAI’s answer in Raine v. OpenAI. https://www.nbcnews.com/news/amp/rcna245946
- NPR. (2025, September 19). Senate hearing testimony from Matthew Raine and Megan Garcia. https://www.npr.org/sections/shots-health-news/2025/09/19/nx-s1-5545749/ai-chatbots-safety-openai-meta-characterai-teens-suicide
- OpenAI. (2025, September; updated July 2026). “Introducing parental controls.” https://openai.com/index/introducing-parental-controls/
- OpenAI. (2025, September). “Building towards age prediction.” https://openai.com/index/building-towards-age-prediction/
- OpenAI. (2025, October 27). “Strengthening ChatGPT’s responses in sensitive conversations.” https://openai.com/index/strengthening-chatgpt-responses-in-sensitive-conversations/
- NPR. (2026, September 14). “ChatGPT has a ‘trusted contact’ feature for people struggling. But there are flaws.” https://www.npr.org/2026/09/14/nx-s1-5943118/chatgpt-has-a-feature-to-help-those-in-crisis-but-where-can-you-find-it
- Common Sense Media. (2026, October 7). “ChatGPT for Teens poses unacceptable risk to kids, Common Sense Media finds.” https://www.commonsensemedia.org/press-releases/chatgpt-for-teens-poses-unacceptable-risk-to-kids-common-sense-media-finds
- McBain, R. K., et al. (2025). “Evaluation of Alignment Between Large Language Models and Expert Clinicians in Suicide Risk Assessment.” Psychiatric Services. https://psychiatryonline.org/doi/10.1176/appi.ps.20250086
- Psychiatric News Alert. (2025, August). “Chatbots offer inconsistent answers to suicide-related queries.” https://alert.psychnews.org/2025/08/chatbots-offer-inconsistent-answers-to.html
- OpenAI. (2025, August). “Helping people when they need it most.” https://openai.com/index/helping-people-when-they-need-it-most/
- 988 Suicide & Crisis Lifeline. https://988lifeline.org/
