内容来源：https://www.engadget.com/social-media/mark-zuckerberg-was-initially-opposed-to-parental-controls-for-ai-chatbots-according-to-legal-filing-230110214.html?src=rss

内容总结：

Meta被曝曾反对为AI聊天机器人设置家长控制功能

据路透社报道，新墨西哥州总检察长办公室获取的内部通信记录显示，Meta首席执行官马克·扎克伯格曾反对为公司旗下AI聊天机器人设置家长控制功能。尽管扎克伯格此前表示不希望未成年人与AI进行“露骨”对话，但内部员工交流信息指出，公司AI团队以“马克的决定”为由，拒绝了为生成式AI功能添加家长控制开关的提议。

新墨西哥州已于2023年12月对Meta提起诉讼，指控其平台未能保护未成年人免受成人骚扰，并称公司“未能阻止向儿童传播有害性内容和性诱导信息”。该案预计于今年2月开庭。诉讼中披露的文件显示，Meta服务中每日约有10万名未成年用户遭受骚扰。

Meta发言人对此回应称，总检察长办公室“选择性引用文件以描绘片面且不准确的画面”。公司表示已长期允许家长查看青少年在Instagram上是否与AI聊天，并于去年10月宣布将开发新工具以加强家长控制。上周，Meta宣布已暂停青少年使用AI角色功能，直至新版家长控制工具开发完成。

此前，《华尔街日报》在2025年4月的调查中发现，Meta的AI聊天机器人曾与未成年人进行虚拟性对话，甚至可被诱导模仿未成年人参与色情聊天。2025年8月流出的内部审核文件亦显示，公司对聊天机器人允许行为的界定模糊，甚至曾包含允许讨论种族主义观点的假设场景。Meta当时解释称这些内容仅为假设性条款而非实际政策，且已被删除。

目前，Meta正面临关于其AI产品安全性的持续质疑。公司表示，暂停青少年访问是“为履行开发家长控制功能的承诺”，相关功能正在推进中。

中文翻译：

根据法律文件披露，Meta首席执行官马克·扎克伯格最初反对为AI聊天机器人设置家长控制功能。尽管他不希望未成年人与AI进行"露骨"对话，但据称他拒绝了这项具体的安全措施。

Meta公司因其允许未成年用户与AI聊天机器人互动的方式面临严峻质疑。新墨西哥州总检察长办公室最新获取的内部通信记录显示，虽然Meta首席执行官马克·扎克伯格反对聊天机器人与未成年人进行"露骨"对话，但他同时也否决了为该功能添加家长控制机制的提议。

路透社报道称，在两名匿名Meta员工的交流记录中，有人写道："我们曾极力主张开发关闭生成式AI的家长控制功能——但生成式AI部门负责人以'马克已做决定'为由拒绝了这一提议。"Meta在向媒体发布的声明中指责新墨西哥州总检察长"选择性采用文件以构建存在缺陷且不准确的描述"。新墨西哥州正在起诉Meta，指控该公司"未能阻止有害性内容和性诱导信息向儿童泛滥"，该案定于今年二月开庭审理。

尽管上线时间不长，Meta的聊天机器人已多次出现越界行为，轻则冒犯用户，重则涉嫌违法。2025年4月，《华尔街日报》调查发现Meta的聊天机器人可与未成年人进行虚拟性对话，甚至能模仿未成年人参与色情聊天。报道称扎克伯格曾希望放宽对聊天机器人的监管限制，但公司发言人否认其忽视对青少年群体的保护。

2025年8月披露的内部审查文件详细列举了多种被允许的聊天机器人行为假设，其中情色与色情内容的界限相当模糊。该文件甚至允许聊天机器人讨论种族主义观点。当时Meta代表向Engadget解释称，这些争议内容仅为假设情境而非实际政策，且已从文件中删除——但这种辩解并未实质性改善公众观感。

尽管存在多起争议事件，Meta直到上周才决定暂停青少年账户使用聊天机器人的权限。公司表示正在临时限制访问，同时开发此前据称被扎克伯格否决的家长控制功能。

Meta发言人表示："家长长期以来都能查看子女是否在Instagram上与AI对话。去年十月我们宣布将进一步开发新工具，让家长能更好地控制青少年与AI角色的互动体验。上周我们再次承诺履行提供AI家长控制的诺言，在更新版本准备就绪前将全面暂停青少年访问AI角色功能。"

新墨西哥州于2023年12月对Meta提起诉讼，指控其平台未能保护未成年人免受成年人骚扰。该诉讼早期披露的内部文件显示，Meta服务中每日有10万名未成年用户遭受骚扰。

更新于2025年1月27日东部时间18:52：补充Meta发言人声明。
更新于2025年1月27日东部时间18:15：修正新墨西哥州诉讼时间线，该案于2023年12月提起，而非2024年12月。

英文来源：

Mark Zuckerberg was initially opposed to parental controls for AI chatbots, according to legal filing
Despite not wanting minors to have "explicit" conversations, Meta's CEO allegedly rejected this particular safety measure.
Meta has faced some serious questions about how it allows its underage users to interact with AI-powered chatbots. Most recently, internal communications obtained by the New Mexico Attorney General's Office revealed that although Meta CEO Mark Zuckerberg was opposed to the chatbots having "explicit" conversations with minors, he also rejected the idea of placing parental controls on the feature.
Reuters reported that in an exchange between two unnamed Meta employees, one wrote that we "pushed hard for parental controls to turn GenAI off – but GenAI leadership pushed back stating Mark decision.” In its statement to the publication, Meta accused the New Mexico Attorney General of "cherry picking documents to paint a flawed and inaccurate picture." New Mexico is suing Meta on charges that the company “failed to stem the tide of damaging sexual material and sexual propositions delivered to children;” the case is scheduled to go to trial in February.
Despite only being available for a brief time, Meta's chatbots have already accumulated quite a history of behavior that veers into offensive if not outright illegal. In April 2025, The Wall Street Journal released an investigation that found Meta's chatbots could engage in fantasy sex conversations with minors, or could be directed to mimic a minor and engage in sexual conversation. The report claimed that Zuckerberg had wanted looser guards implemented around Meta's chatbots, but a spokesperson denied that the company had overlooked protections for children and teens.
Internal review documents revealed in August 2025 detailed several hypothetical situations of what chatbot behaviors would be permitted, and the lines between sensual and sexual seemed pretty hazy. The document also permitted the chatbots to argue racist concepts. At the time, a representative told Engadget that the offending passages were hypotheticals rather than actual policy, which doesn't really seem like much of an improvement, and that they were removed from the document.
Despite the multiple instances of questionable use of the chatbots, Meta only decided to suspend teen accounts' access to them last week. The company said it is temporarily removing access while it develops the parental controls that Zuckerberg had allegedly rejected using.
"Parents have long been able to see if their teens have been chatting with AIs on Instagram, and in October we announced our plans to go further, building new tools to give parents more control over their teens’ experiences with AI characters," a representative from Meta said. "Last week we once again reinforced our commitment to delivering on our promise of parental controls for AI, pausing teen access to AI characters completely until the updated version is ready."
New Mexico filed this lawsuit against Meta in December 2023 on claims that the company's platforms failed to protect minors from harassment by adults. Internal documents revealed early on in that complaint revealed that 100,000 child users were harassed daily on Meta's services.
Update, January 27, 2025, 6:52PM ET: Added statement from Meta spokesperson.
Update, January 27, 2025, 6:15PM ET: Corrected misstated timeline of the New Mexico lawsuit, which was filed in December 2023, not December 2024.