OpenAI and Anthropic are introducing new ways to detect underage users. Since OpenAI updated its guidelines on how ChatGPT should interact with users aged 13 to 17, Anthropic is working on a new way to identify and boot users under 18.
On Thursday, OpenAI announced it chatgpt Model Specification – Guidelines for how its chatbots should behave will include four new principles for users under 18. Now, its goal is to keep ChatGPT “putting the safety of teens first, even if it conflicts with other goals.” This means guiding teens toward safer options when other user interests, such as “maximum intellectual freedom”, conflict with safety concerns.
It also states that ChatGPT should “promote real-world support”, including encouraging offline relationships, as well as how ChatGPT should set clear expectations when interacting with young users. The model spec states that ChatGPT should “treat teens like teens” by providing “warmth and respect” rather than giving condescending answers or treating teens like adults.
OpenAI says the update to ChatGPT’s model spec should result in “stronger guardrails, safer options, and incentives to seek reliable offline support when conversations move into high-risk territory.” The company says ChatGPT will urge teens to contact emergency services or crisis resources if they see signs of “imminent risk.”
With this change, OpenAI says it is in the “early stages” of launching an age prediction model that will attempt to predict someone’s age. If it detects that a person may be under 18, OpenAI will automatically apply teen protection measures. It will also give adults a chance to verify their age if they have been incorrectly flagged by the system.
Anthropic, which does not allow users under the age of 18 to chat with Cloud implement measures This will be used to detect and disable accounts of underage users. It is developing a new system capable of detecting “subtle conversation signals that a user may be underage” and says it already flags users who identify themselves as minors during chat.
Anthropic also outlines how it trains the cloud to respond to signals about suicide and self-harm, as well as its progress in reducing cynicism, which can confirm harmful thinking. The company says its latest models have “the least chatter of all models to date”, with the Haiku 4.5 performing the best, as it corrected its chatter behavior 37 percent of the time.
“At face value, this evaluation shows that there is significant room for improvement across all of our models,” says Anthropic. “We think the results reflect a trade-off between idealized warmth or friendliness, on the one hand, and sycophancy, on the other.”
Update, December 18: Clarified that Anthropic does not allow users under the age of 18 to use the cloud.
