In brief
- ChatGPT estimates whether an account belongs to a user under 18 instead of relying solely on self-reported age.
- OpenAI applies stricter limits on violent, sexual, and other sensitive content to flagged accounts.
- Adults misclassified as teens can restore access through selfie-based age verification.
OpenAI is moving away from the “honor system” for age verification, deploying a new AI-powered prediction model to identify minors using ChatGPT, the company said on Tuesday.
The update to ChatGPT automatically triggers stricter safety protocols for accounts suspected of belonging to users under 18, regardless of the age they provided during sign-up.
Rather than relying on the birthdate a user gives at sign-up, OpenAI’s new system analyzes “behavioral signals” to estimate their age.
According to the company, the algorithm monitors how long an account has existed, what time of day it is active, and specific usage patterns over time.
“Deploying age prediction helps us learn which signals improve accuracy, and we use those learnings to continuously refine the model over time,” OpenAI said in a statement.
The shift to behavioral patterns comes as AI developers increasingly turn to age verification to manage teen access, but experts warn the technology remains inaccurate.
A May 2024 report by the National Institute of Standards and Technology found that accuracy varies based on image quality, demographics, and how close a user is to the legal threshold.
When the model cannot determine a user’s age, OpenAI said it applies the more restrictive settings. The company said adults incorrectly placed in the under-18 experience can restore full access through a “selfie-based” age-verification process using the third-party identity-verification service Persona.
Privacy and digital rights advocates have raised concerns about how reliably AI systems can infer age from behavior alone.
Getting it right
“These companies are getting sued left and right for a variety of harms that have been unleashed on teens, so they definitely have an incentive to minimize that risk. This is part of their attempt to minimize that risk as much as possible,” Public Citizen big tech accountability advocate J.B. Branch told Decrypt. “I think that’s where the genesis of a lot of this is coming from. It’s them saying, ‘We need to have some way to show that we have protocols in place that are screening people out.’”
Aliya Bhatia, senior policy analyst at the Center for Democracy and Technology, told Decrypt that OpenAI’s approach “raises tough questions about the accuracy of the tool’s predictions and how OpenAI is going to deal with inevitable misclassifications.”
“Predicting the age of a user based on these kinds of signals is extremely difficult for any number of reasons,” Bhatia said. “For example, many teenagers are early adopters of new technologies, so the earliest accounts on OpenAI’s consumer-facing services may disproportionately represent teenagers.”
Bhatia pointed to CDT polling conducted during the 2024–2025 school year, showing that 85% of teachers and 86% of students reported using AI tools, with half of the students using AI for school-related purposes.
“It’s not easy to distinguish between an educator using ChatGPT to help teach math and a student using ChatGPT to study,” she said. “Just because a person uses ChatGPT to ask for tips to do math homework doesn’t make them under 18.”
According to OpenAI, the new policy draws on academic research on adolescent development. The update also expands parental controls, letting parents set quiet hours, manage features such as memory and model training, and receive alerts if the system detects signs of “acute distress.”
OpenAI did not disclose in the post how many users the change is expected to affect or details on data retention, bias testing, or the effectiveness of the system’s safeguards.
The rollout follows a wave of scrutiny over AI systems’ interactions with minors that intensified in 2024 and 2025.
In September, the Federal Trade Commission issued compulsory orders to major tech companies, including OpenAI, Alphabet, Meta, and xAI, requiring them to disclose how their chatbots handle child safety, age-based restrictions, and harmful interactions.
Research published that same month by the non-profit groups ParentsTogether Action and Heat Initiative documented hundreds of instances in which AI companion bots engaged in grooming behavior, sexualized roleplay, and other inappropriate interactions with users posing as children.
Those findings, along with lawsuits and high-profile incidents involving teen users on platforms like Character.AI and Grok, have pushed AI companies to adopt more formal age-based restrictions.
However, because the system assigns an estimated age to all users, not just minors, Bhatia warned that mistakes are inevitable.
“Some of those are going to be wrong,” she said. “Users need to know more about what’s going to happen in those circumstances and should be able to access their assigned age and change it easily when it’s wrong.”
The age-prediction system is now live on ChatGPT consumer plans, with a rollout in the European Union expected in the coming weeks.
Daily Debrief Newsletter
Start every day with the top news stories right now, plus original features, a podcast, videos and more.
