OpenAI rolls out age prediction system for ChatGPT to protect under-18 users

ChatGPT

• OpenAI launches age prediction model to protect minors on ChatGPT
• Safeguards limit sensitive content for users likely under 18
• Users wrongly flagged as minors can verify age via Persona for full access

Did our AI summary help?

OpenAI said on Tuesday that it is rolling out an age prediction model across ChatGPT’s consumer plans, as part of broader efforts to better protect younger users on the platform.

According to the company, the model uses a mix of account-level and behavioural signals to estimate whether an account belongs to someone under the age of 18. These signals include usage patterns over time, how long an account has existed, typical activity hours, and a user’s stated age.

If the system determines that a user is likely under 18, ChatGPT will automatically apply safeguards designed to limit exposure to sensitive material. This includes content related to self-harm and other topics considered inappropriate for minors.

The rollout follows a series of safety-focused updates from OpenAI, as the company faces growing scrutiny over how artificial intelligence tools affect children and teenagers. OpenAI, along with other major tech firms, is currently under investigation by the US Federal Trade Commission over whether AI chatbots can negatively impact young users.

The company is also named in several wrongful death lawsuits, including one involving the suicide of a teenage boy, which has intensified pressure on OpenAI to strengthen its protections.

OpenAI said users who are incorrectly flagged as being under 18 will be able to regain full access by verifying their age through Persona, a third-party identity verification service. Persona is already used by several tech companies, including Roblox, which has faced similar calls from lawmakers to improve child safety on its platform.

The age prediction system builds on earlier commitments made by OpenAI last year. In August, the company said it would introduce parental controls to help guardians understand and manage how their teenagers use ChatGPT. Those controls were rolled out the following month, alongside confirmation that OpenAI was developing an age detection system.

In October, OpenAI also convened an advisory council of eight experts to study how AI systems may affect users’ mental health, emotions, and motivation.

OpenAI said it will continue refining the accuracy of the age prediction model over time. The feature is expected to roll out in the European Union in the coming weeks, with adjustments made to meet regional regulatory requirements.