OpenAI to implement proactive safety measures in ChatGPT, including conversation termination and bans

OpenAI has announced forthcoming updates to ChatGPT that introduce stricter safeguards against violent content. These new protocols allow the model to autonomously terminate a conversation if it repeatedly focuses on violence or serious harm to others. Furthermore, repeated violations by the same account will result in an outright ban. This enhancement moves beyond merely refusing to answer dangerous prompts; it introduces an active interruption mechanism. The updated rules apply universally, affecting all users. For discussions concerning historical attack incidents, the platform will now limit or refuse to provide details regarding weapons, equipment, or acquisition methods, while also restricting side-by-side comparisons of casualty figures. While adult users face these content limitations, accounts identified as minors will receive an additional layer of protection, with practical information on weapons and security systems being blocked even if the query appears innocuous. OpenAI will deploy an improved age prediction system, utilizing various user behavior signals to automatically apply these teen protections, allowing users to submit identity verification if misclassified.