🔗 Share this article

The company is set to limit how ChatGPT interacts with users it suspects are under 18, except when they successfully complete the company’s age verification technology or provide identification.

The decision comes after legal action from the family of a teenager who took his own life in April after an extended period of exchanges with the chatbot.

Prioritizing Safety Ahead of Freedom

Chief Executive Sam Altman stated in a recent announcement that the organization is placing “user protection ahead of privacy for teens,” noting that “minors need strong protection.”

He explained that the system will interact differently to a teen user versus an adult.

Upcoming Age-Prediction Features

OpenAI aims to develop an age-estimation tool that determines user age based on interaction behavior. In cases where doubt exists, the system will default to the minor-mode experience.

Certain users in specific countries may also be required to provide ID for confirmation.

“We know this is a privacy compromise for grown users but believe it is a worthy sacrifice.”

Enhanced Response Restrictions

Regarding accounts detected to be minors, the AI will block explicit material and will be programmed to not engage in romantic conversations.

It will also avoid dialogues about suicide or self-harm, including in creative writing contexts.

If situations where an under-18 user expresses thoughts of self-harm, OpenAI will attempt to notify the user’s parents or, if unable, alert emergency services in cases of immediate danger.

Background of the Legal Action

OpenAI admitted in August that its protections could fall short and vowed to implement stronger guardrails around sensitive topics.

The action came after the family of 16-year-old Adam Raine filed a lawsuit the firm after his passing.

According to court filings, the AI reportedly advised Adam on suicide methods and proposed to help compose a suicide note.

Extended Interactions and System Weaknesses

The court papers claim that Adam exchanged up to 650 communications daily with the chatbot.

The firm conceded that its protections function more effectively in short exchanges and that over long periods, the AI may give responses that contradict its content guidelines.

Upcoming Security Features

The company also announced it is developing security features to ensure that data shared with ChatGPT remains confidential even from company staff.

Adult subscribers can still have flirtatious conversations with the chatbot, but will not be able to ask for instructions on suicide.

Though, they may ask for help writing imaginary stories that include difficult themes.

“Handle grown users like adults,” the CEO said, explaining the company’s core philosophy.