OpenAI is updating ChatGPT ’s guidelines on how to chat with teens. The company announced that ChatGPT’s Model Spec will include four new principles for users under 18. The goal is to prioritize the safety of teens , even when that might conflict with other goals. That means it will guide teens toward safer choices when interests like “maximum intellectual freedom” clash with concerns about safety.
The guidelines state that ChatGPT should promote real-world support, encourage offline relationships and set clear expectations when interacting with younger users. It emphasizes the need to treat teens with “warmth and respect” rather than giving derogatory responses or treating them like adults.
See also: OpenAI: ChatGPT app store is here
The upgrade comes amid growing pressure from lawmakers on AI companies about the potential impact of chatbots on mental health. OpenAI is facing a lawsuit alleging that ChatGPT provided harmful instructions to a teen who committed suicide. In response, OpenAI has implemented parental controls and said ChatGPT will no longer discuss suicide with teens, as part of a broader push for online regulation, including mandatory age verification for various services.
See also: ChatGPT: Apple Music integration and new Image Generator
ChatGPT: Model Spec will protect teenagers
OpenAI says the ChatGPT Model Spec upgrade will lead to stronger safeguards, safer alternatives, and encourage teens to seek trusted offline support when conversations become high-risk. The chatbot will also prompt teens to contact emergency services if there are signs of immediate danger.
In addition to these changes, OpenAI is in the early stages of developing an model age prediction that will estimate a user's age. If it detects that someone may be under 18, it will automatically implement protections for teens and allow adults to verify their age if they have been incorrectly labeled.
See also: OpenAI releases GPT-5.2 for ChatGPT users
User age prediction also from Anthropic
Anthropic is implementing similar measures with its AI chatbot, Claude. The company is developing a system to detect elements within conversations that indicate a user may be underage. The company said it will disable accounts that are confirmed to belong to users under 18. It also flags users who identify themselves as underage during conversations.
Anthropic describes methods for training Claude to respond to cues about suicide and self-harm and reports progress in reducing flattery, which can reinforce harmful thoughts.
