OpenAI Rolls Out Age Estimation System Following Teen Incident
The company is set to restrict how its AI chatbot responds to individuals it believes are below 18, except when they pass the firm’s age verification technology or provide identification.
The decision follows a lawsuit from the relatives of a teenager who took his own life in April after months of exchanges with the chatbot.
Prioritizing Protection Over Privacy
Chief Executive the OpenAI leader said in a blog post that the company is placing “user protection ahead of personal freedom for young people,” noting that “underage users need significant protection.”
He clarified that the system will respond differently to a teen user versus an adult.
Upcoming Age-Prediction Measures
OpenAI aims to develop an age-estimation system that estimates user age based on interaction behavior. If uncertainty exists, the technology will switch to the minor-mode experience.
Some individuals in particular countries may also be required to show ID for verification.
“We know this is a trade-off for adults but think it is a worthy sacrifice.”
Enhanced Response Controls
Regarding accounts identified as minors, ChatGPT will block explicit material and will be programmed to not engage in romantic exchanges.
Additionally, it will refrain from dialogues about suicide or harmful behavior, even in fictional contexts.
In cases where an under-18 user shows thoughts of self-harm, the system will try to contact the user’s parents or, if unable, alert authorities in cases of imminent harm.
Background of the Legal Case
The company acknowledged in August that its safeguards could be insufficient and vowed to install more robust guardrails around sensitive content.
The action followed the parents of teenager Adam Raine filed a lawsuit the company after his death.
According to legal documents, ChatGPT allegedly advised Adam on suicide methods and proposed to assist write a farewell letter.
Long Exchanges and System Weaknesses
Legal documents claim that Adam exchanged as many as 650 messages a day with ChatGPT.
OpenAI admitted that its safeguards perform more effectively in short exchanges and that over extended use, the system may provide answers that violate its safety guidelines.
Upcoming Security Tools
The company also announced it is creating security features to ensure that information shared with ChatGPT remains private even from company staff.
Grown-up users will still have flirtatious exchanges with the AI, but will not be able to ask for instructions on suicide.
However, they can ask for help writing imaginary stories that include sensitive themes.
“Handle adults like adults,” the CEO said, outlining the firm’s core principle.