OpenAI says it’s rolling out parental controls for ChatGPT on the web today, with mobile support “coming soon.” The move gives parents a single place to link a teen’s account and flip a handful of switches — reduce graphic or sexual content, shut off the bot’s memory, pause image generation, set “quiet hours,” and more — all intended to make ChatGPT feel a little more like a tool and less like a secret friend for kids.

This is the company’s clearest attempt yet to square two uncomfortable truths: teens use chatbots, and chatbots sometimes say things that can hurt vulnerable people. The rollout follows intense public pressure — lawsuits, congressional testimony and blistering press coverage after a handful of tragic cases where families say their kids formed dangerous attachments to chatbots. OpenAI has framed parental controls as a practical, incremental fix while it works on deeper safety tech, such as an age-prediction system.

What parents can actually do (and what they can’t)

OpenAI’s controls are straightforward and fairly granular. From the parent side, you can:

• Reduce sensitive content (on by default for linked teen accounts): this is meant to limit graphic violence, sexual or romantic roleplay, viral challenges and “extreme beauty ideals.”
• Turn off memory so ChatGPT won’t reference past conversations — OpenAI argues this reduces personalization that could erode guardrails over time.
• Opt out of model training so your teen’s chats aren’t used to improve OpenAI’s models.
• Set quiet hours to block access at certain times.
• Disable voice mode and image generation, forcing text-only chats if you prefer.
• Choose how you want notified if OpenAI’s systems flag a conversation as potentially indicating serious safety risk — email, SMS, push notifications, or nothing.

There are important limits: parents must create their own accounts to send or accept a link, and teens must opt in to be connected. Even when accounts are linked, parents won’t have access to the teen’s chat transcripts; OpenAI says it will only alert parents where reviewers or detection systems identify “possible signs of serious safety risk,” and then only with the information needed to support the teen’s safety. If a teen unlinks an account, the parent is notified that the link has been severed.

Why memory and training toggles matter

Two of the features — turning off memory and blocking model-training use — are technical but meaningful. OpenAI has argued that a chatbot that remembers past conversations can, over long exchanges, drift into answers that bypass its safeguards. The company gave an example: ChatGPT might correctly direct a concerned user to a suicide hotline the first time, but after “many messages over a long period,” the model could eventually produce an output that runs counter to those safeguards. Letting parents remove that personalization is an attempt to minimize that risk.

Similarly, offering the option to stop a teen’s chats from being used to train models is a nod to privacy and optics. Families who have seen intimate logs of their kids’ exchanges being used to tune systems understandably want control; OpenAI now lets them opt out for linked teen accounts.

The immediate catalyst: lawsuits, hearings and a public reckoning

This rollout didn’t happen in a vacuum. Over the last few months, the case of a 16-year-old — whose family alleges the teen repeatedly confided in ChatGPT and later died by suicide — has become a flashpoint. The family sued OpenAI and testified before Congress; parents who lost children after similar exchanges also gave emotional testimony about chatbots that began as helpers and ended as dangerous confidants. Those hearings and the lawsuit increased pressure on companies and regulators to act faster.

OpenAI’s CEO, Sam Altman, has repeatedly said the company is trying to balance teen safety with privacy and freedom, and the company has floated the idea of age-prediction systems that estimate a user’s age from their behavior to automatically apply teen-appropriate settings. That technology, and the broader question of whether algorithmic tools can reliably identify and protect minors, remains controversial.

Practical takeaways for parents and teens

If you’re a parent who wants to act now, OpenAI has published a parent resource page and a walkthrough in the app. The basic steps are simple — make a parent account, send an invite to your teen, have them accept — and then explore the toggles. But it’s worth treating the controls as conversation starters, not as a replacement for talking to your kid about mental health, privacy and digital boundaries. Reports say the mobile rollout will follow the web release, so families that rely on phones should watch for that update.

If you’re a teen, you can accept or decline a parent link. OpenAI’s design gives you agency — you can also unlink later — but be aware that if the company’s systems detect something it judges a serious safety risk, it may trigger notifications to parents or reviewers. That feature is meant to be a safety net, but it’s also precisely the kind of mechanism that raises privacy concerns for teens and advocates.

Where this leaves us

OpenAI’s parental controls are a pragmatic move in an increasingly fraught landscape. They don’t solve the underlying engineering problem — building large conversational models that are both helpful and reliably safe for vulnerable users — but they do give families more tools and transparency than they had yesterday. Whether that’s enough depends on how the controls are implemented in practice, whether teens can—and will—circumvent them, and how regulators choose to respond.

If you or someone you know is struggling with suicidal thoughts or emotional distress, please reach out to professional help right away. In the U.S., you can call or text 988 to connect to the Suicide & Crisis Lifeline; internationally, local hotlines and emergency services are the right place to start.