In the past users had to rein themselves in; now the AI grabs the microphone first. OpenAI recently announced that over the coming weeks it will add a harder gate against violent content to ChatGPT: when a conversation keeps circling around violence or thoughts of seriously harming others, the model may end the conversation on its own; if the same account is repeatedly cut off over such issues, a ban will also follow. This is not rule-making from scratch—ChatGPT had already prohibited providing guidance on planning or carrying out violence; this upgrade adds "proactive cutting off" to the enforcement process, moving from "passive refusal" to "active interruption," with the platform's handling logic for dangerous content clearly moved forward.
Same rules for all age groups
The new rules extend the restrictions to all users, with no leniency for being an adult. The adjustment focuses on the standard of answers about "historical attack incidents"—the model will reduce or even refuse to provide details from a safety perspective, covering the weapons, equipment, and means of acquisition used in attacks, and will also limit horizontal comparisons of casualty figures that could amplify the perpetrator's notoriety. In other words, adults and minors face the same tightened boundaries; the difference is only in depth: adult users will likewise be stonewalled, just without an additional block on practical information about weapons and security.
Minors get an extra layer
For accounts identified as belonging to users under 18 years old, OpenAI will additionally block practical information on weapons and security systems that could be used to carry out violence, effective even when the question shows no visible violent intent. Teenagers are still allowed to seek help from ChatGPT—how to prevent or de-escalate violence, or contacting a trusted adult, are not on the blocked list. An accompanying improved age predictor will combine multiple signals such as user behavior to continuously determine whether an account may belong to a minor, automatically applying teen protections when necessary; if misjudged, users can proactively submit identity information for verification.