OpenAI tightens ChatGPT limits on weapons-related information for teens, even for harmless-looking questions

OpenAI said on Friday it is tightening limits on the weapons-related information ChatGPT provides to teens, even when a question appears harmless, according to Reuters. The company framed the change as part of new safeguards.

What was announced

Per the Reuters report, the restriction applies to weapons-related content served to teenage users of ChatGPT. It covers questions that look benign on their face, not only those that are overtly harmful.

The Reuters URL describes the effort as a violence prevention push.

This is a policy and safeguard change to an existing product. No new model, version, pricing, context window or benchmark figures were attached to the announcement.

What this means

The reported change shifts the standard for teen accounts from judging whether a question looks harmful to restricting the weapons topic area more broadly. That matters for anyone building on or evaluating ChatGPT's behavior with younger users: responses on weapons subjects may be more restricted than before, including for questions that previously would have passed as ordinary curiosity. Specific details of the new safeguards were not available in the Reuters summary.