OpenAI prioritizes teen safety in ChatGPT

OpenAI has decided that when privacy, freedom, and safety come into conflict, protecting people under 18 will take priority in ChatGPT. To apply this policy, it is developing a system that estimates each user’s age based on how they use the service.

ChatGPT is intended for people 13 and older. If the system cannot determine with sufficient confidence whether someone is an adult, OpenAI says it will default to the experience designed for minors. In some countries or situations, it may also request an identity document.

The company acknowledges that age verification may reduce adult users’ privacy, but considers that an acceptable cost for protecting teenagers.

Three principles that clash

OpenAI explains its decision through three goals that cannot always be achieved at the same time.

The first is privacy. The company says conversations with an AI can contain information as personal as what someone shares with a doctor or lawyer. That is why it is developing safety features to keep this data protected, including from its own employees.

That protection would not be absolute. Automated systems would review possible serious misuse, and some cases could be referred for human review, for example:

• Threats against someone’s life.
• Plans to harm other people.
• Large-scale risks, such as a potential mass cyberattack.

OpenAI also says it is arguing to policymakers that sensitive conversations with an AI should receive a level of protection similar to that given to other private communications.

The second principle is adult freedom. The company wants adults to be able to use its tools broadly as long as they do not cause harm or restrict other people’s freedom. For example, an adult might request a flirtatious conversation, even though that would not be the model’s default behavior.

The distinction would also apply to sensitive topics. The system should not explain how to commit suicide, but it could help an adult write a work of fiction that includes a suicide. OpenAI’s internal approach is to treat adults like adults and expand their options as models become better at following instructions.

Stricter rules for teenagers

For minors, OpenAI is proposing a different limit: safety will take priority over privacy and freedom. ChatGPT should not engage in flirtatious conversations with a teenager, even if they request them, or discuss suicide or self-harm as part of a creative writing exercise.

If a user under 18 shows signs of suicidal ideation, OpenAI says it will try to contact their parents. If it cannot do so and believes there is an imminent danger, it will also contact the authorities.

To make these rules work, the company is combining its age prediction system with new parental controls. It has not presented this measure as a perfect solution: it acknowledges that the system can make mistakes and that identifying someone’s age without requesting documents creates technical and privacy challenges.

For you, the change will depend on how ChatGPT classifies your account. An adult could face an identity check or a more limited experience if the system is unsure of their age. A teenager will face more restrictions in romantic conversations and content related to self-harm, even when requesting it for creative purposes.

The decision highlights the central problem facing AI assistants: the more personal their conversations become, the harder it is to protect privacy without ignoring signs of danger. OpenAI has chosen a stricter line for minors, but it will still be necessary to monitor how age prediction works, when human intervention is triggered, and what safeguards are in place to prevent mistakes.