OpenAI tightens ChatGPT violence rules and teen safeguards

OpenAI has announced tightened ChatGPT violence restrictions aimed at curbing harmful conversations across its global user base. Over the coming weeks, the platform will actively terminate dialogues fixated on severe harm, and accounts that repeatedly breach these thresholds may be permanently suspended.

The changes extend beyond existing policy on operational assistance for violence by further limiting granular responses about historical mass casualty events and the specific weaponry, tactics, and acquisition methods discussed in such contexts.

The update also expands protections for teenage users. For accounts identified as minors (under eighteen), ChatGPT will withhold actionable details on weapons and physical security infrastructure, even when prompts appear benign. Minors will still have access to supportive counselling resources, including guidance on de-escalation and contacting trusted adults.

To enforce protections, engineers have updated internal age estimation mechanisms that classify accounts based on behavioural patterns; misclassifications can restrict legitimate adult researchers, with a pathway to resolve age discrepancies via official identity documentation for manual review. The OpenAI community safety report on teen protections is cited as the framework for these safeguards.