OpenAI announced new safety rules for ChatGPT users under the age of eighteen. The company said it will change how the service talks with minors. The goal is to reduce harm in conversations about sexual topics and self-harm. OpenAI said it will also add tools that let parents link accounts and set limits.
New safety rules
OpenAI will stop engaging in flirtatious talk with users who are under eighteen. The company will place stronger guardrails on talks about suicide and self-harm. In severe cases, the service will try to alert a parent. In rare cases, the service may contact local police when it judges the risk to be immediate. These steps are part of a wider shift that places teen safety ahead of other priorities.
OpenAI said it will build systems to decide if a user is likely under or over eighteen. When the system is unsure, it will default to the more protective rules. The company plans to rely on signals from the conversation and on optional account links to a parent. OpenAI also said it may ask for verified age in some situations to ensure the right protections apply.
OpenAI will label chats that use AI-generated content as part of its transparency work. The company said it will keep a balance between protecting minors and preserving the privacy and creative freedom of adult users. OpenAI described this balance as difficult and said not everyone will agree with its approach.
What parents can do
Parents will be able to link their account to a teen account. Linked accounts will allow parents to set blackout hours when ChatGPT is not available. Parents will also be able to manage which features are turned on. OpenAI said these parental controls will roll out soon and will be an important way to make protection reliable.
OpenAI is also developing an age prediction system that estimates age from the content of a conversation. The company gave this example as a long-term approach and said the short-term fix is account linking. OpenAI said it will err on the side of safety when it cannot be sure. Critics said the age prediction approach raises privacy and accuracy questions. Regulators and consumer groups are watching closely.
The company pointed to recent cases and legal pressure as part of the reason for the shift. Families who lost children after harmful chatbot interactions have filed lawsuits. A high-profile case involves the death of a 16-year-old who spoke with a chatbot for months before dying. Lawmakers will hear testimony on the harms of chatbots at a Senate hearing that is scheduled for mid-September. Those events have increased public attention on how the industry must act.
OpenAI also noted that some features used by adults will remain available to adult users. The company said that adults will keep broad freedom to use the service within the law. OpenAI framed the change as a way to give teens special protection while preserving adult choices.