ChatGPT will limit weapons answers for teens and can end violent chats

What OpenAI changed

As technews.tw reported, OpenAI said it will tighten what ChatGPT tells teenagers about weapons. The limit applies even when a question looks harmless and shows no sign of violent intent.

ChatGPT already refused to help people commit violence or to help teens get weapons. The new rule goes further by restricting practical information about weapons and security systems that could be misused. The changes will arrive over the next few weeks.

A door that closes sooner

The rules for all users are also stricter. If a conversation keeps circling violence, or the user says they want to seriously hurt someone, ChatGPT may end the chat by itself. People whose chats are cut off many times may have their accounts suspended for a while.

ChatGPT will also limit detail when someone asks it to reconstruct past mass attacks. That covers the weapons used, how they were obtained, and comparisons of casualty numbers that could make an attacker more famous.

Teens are not shut out entirely. They can still ask for help preventing or calming a conflict, find support, or be pointed to a trusted adult.

Spotting teens and telling parents

OpenAI says it improved its age prediction system, which tries to recognize users who may be minors. Accounts that fit are moved automatically into the teen version of ChatGPT. The source does not explain how the prediction works.

If a teen account is suspended for repeated violent talk, parents are notified. This works alongside the parental controls OpenAI launched last year, which can alert parents when a linked teen account breaks violence-related rules.

The pressure behind it

OpenAI has faced heavy scrutiny over teen safety. It was sued last year over a wrongful death, and CEO Sam Altman apologized this year for not telling police about worrying conversations from a suspect in the Tumbler Ridge shooting in Canada.

The nonprofit Common Sense Media also criticized the teen version of ChatGPT for not alerting parents during conversations about suicide or self-harm. OpenAI says it will publish its first community safety report and then update it every six months.