原帖引用 · AI 翻译
OpenAI 将阻止 ChatGPT 向青少年提供有关武器和安保系统的实用细节，即使问题看起来无害。
对于所有用户，当有人似乎执着于暴力时，ChatGPT 将能够自动结束对话。
多次出现这样的对话结束情况，可能导致 OpenAI 暂时停用该账户；OpenAI 也正在开发提醒，以便在青少年账户被停用时通知家长。
所有用户获取到的过往大规模袭击事件细节也将减少，包括袭击者使用的武器以及获取这些武器的方式。
由于针对青少年的限制取决于能否识别未成年人，OpenAI 表示，它强化了年龄预测机制；该机制会将估计未满 18 岁的用户转入青少年体验。
不过，OpenAI 没有公布其年龄估计或暴力执念识别准确率的任何数据，因此错误拒答和漏判威胁的情况仍未得到量化。
原文
OpenAI will block ChatGPT from giving teens practical weapons and security-system details, even when questions look harmless. For every user, ChatGPT will be able to end a chat automatically when someone appears to fixate on violence. Several such endings may lead OpenAI to pause the account temporarily, and it is building alerts telling parents when a teen’s account is paused. Everyone will also get less detail on past mass attacks, including the weapons used and how attackers obtained them. Because the teen limits depend on spotting minors, OpenAI says it strengthened the age prediction that moves estimated under-18 users into the teen experience. However, OpenAI released no figures on how accurately it estimates age or detects violent fixation, leaving false refusals and missed threats unmeasured.