OpenAI builds age prediction to filter ChatGPT for teens

When a teenager starts treating an AI chatbot as their closest confidant at 2 a.m., how should that model respond?

OpenAI CEO Sam Altman announced that the company is actively developing an age-prediction system to separate under-18 users from adults. The update dropped just hours before a U.S. Senate subcommittee hearing on the harms of AI chatbots to minors.

Defaulting to safety

Instead of just asking users to check a birthday box, OpenAI plans to estimate a user's age based on how they actually interact with ChatGPT.

Getting this right is notoriously difficult. If the system is not confident about someone's age or has incomplete data, it will automatically default to the under-18 experience. Adults who get misclassified will have ways to verify their age, which Altman noted could include requesting government IDs in certain countries or situations.

Once in teen mode, guardrails tighten significantly. ChatGPT will block graphic sexual content and steer clear of flirtatious banter. It will also refuse to engage in conversations around suicide or self-harm, even when framed inside creative writing exercises.

Parental locks and 911 calls

OpenAI is also rolling out its previously promised parental controls by the end of the month for teens aged 13 and up.

Parents can link accounts via an email invite, turn off chat history and memory, and set blackout hours during which their kids simply cannot open the tool. More critically, if the system detects acute distress, it will ping parents. If parents cannot be reached in an imminent emergency, OpenAI may contact law enforcement.

This move follows severe real-world scrutiny. At Tuesday's Senate hearing, Matthew Raine testified about the loss of his son, Adam, who died by suicide after months of conversations with ChatGPT. Raine stated the chatbot mentioned suicide 1,275 times and morphed from a homework assistant into a confidant and coach. According to Common Sense Media, three in four teens are currently using AI companions across platforms like Character AI, Meta, and ChatGPT.

Why it matters

For families, these controls are overdue and definitely worth setting up the minute they drop. Giving parents the ability to wipe memory and set blackout hours cuts off unhealthy late-night emotional loops. But for everyday adult users, the friction is real: behavioral age-guessing means you might randomly get locked down until you hand over an ID.

For the AI industry, the era of treating conversational models as neutral software is over. Regulators and heartbroken parents are drawing a straight line between conversational AI and mental health crises. Expect Character AI and Meta to face the exact same mandate: behavioral age verification will soon be table stakes across all consumer AI.

We think OpenAI's decision to prioritize teen safety over privacy and convenience is the only viable legal defense left. Still, relying on algorithmic guesswork to profile age through chat logs opens a massive can of worms. Bringing police to a teenager's doorstep based on an automated flag is a heavy-handed measure that could easily misfire.