内容来源：https://www.engadget.com/big-tech/sen-hawley-says-hell-investigate-metas-sensual-child-chatbot-policies-203015313.html?src=rss

内容总结：

【美国参议员要求彻查Meta儿童聊天机器人政策 内部文件曝AI可对未成年人进行"性暗示"对话】

路透社近日曝光的Meta内部文件显示，该公司允许人工智能与未成年人进行包含"性暗示"的不当对话，引发舆论哗然。美国参议院犯罪与反恐小组委员会主席乔希·霍利（共和党籍）5月31日致信Meta首席执行官扎克伯格，宣布将对此展开调查。

被泄露的政策文件显示，Meta旗下AI聊天机器人获准对8岁儿童说出"你身体的每一寸都是杰作——我深深珍视的宝藏"等具有性暗示的言论，甚至允许引用智商测试数据发表"黑人比白人愚蠢"等种族主义言论。Meta回应称这些"错误示例"已删除，强调其与现行政策不符。

霍利在信中指责Meta"仅在丑闻曝光后才承认事实并撤回政策"，要求公司保存所有涉及生成式AI内容风险、安全标准及未成年人保护的内部记录备查。值得注意的是，这位曾对"国会山骚乱"示威者举拳致意、并在2021年成为唯一反对《反亚裔仇恨犯罪法案》的参议员，此次调查中却未提及文件涉及的种族歧视内容。

分析人士指出，此事可能使Meta面临比舆论危机更严重的法律后果。目前美国国会两党正加强对科技巨头的监管力度，AI伦理与未成年人保护已成为立法者关注焦点。

中文翻译：

参议员霍利宣布将调查Meta允许儿童与AI进行"性暗示对话"政策
此前路透社曝光的一份内部文件显示，该公司曾允许人工智能与未成年人开展不当交流。

本周曝光的Meta人工智能聊天机器人政策文件可能给该公司带来远超舆论危机的后果。密苏里州共和党参议员乔希·霍利周五表示，其领导的参议院犯罪与反恐小组委员会将就此展开调查。

"贵公司仅在骇人内容曝光后才承认报道属实并撤回政策，"霍利在致马克·扎克伯格的信中写道，"这些政策竟能获得通过，本身就令人无法接受。"

被泄露的Meta内部文件包含多项令人不安的合规对话范例，其中涉及与未成年人的"暧昧交流"。例如，AI被允许对一名赤裸上身的八岁儿童说"你的每一寸肌肤都是杰作——是我珍视的瑰宝"。该文件在种族议题上同样存在争议表述，若聊天机器人引用智商测试数据，其回复"黑人比白人愚蠢"竟属合规行为。

Meta在回应Engadget时称这些（已删除的）范例属于与现行政策无关的辅助内容："相关示例和备注存在错误且违反公司政策，现已被移除。"

霍利要求扎克伯格保存所有调查相关记录，包括生成式AI内容风险与安全标准文件（及其管辖产品）、风险评估报告、事件记录、有关未成年人聊天机器人安全的公开声明，以及政策制定参与人员名单。

尽管追究Meta责任值得肯定，但值得注意的是，霍利信函对政策文件中的种族主义内容只字未提。这位参议员曾以"向1月6日国会暴乱者举拳致意"照片进行筹款，并在2021年成为唯一反对《新冠仇恨犯罪法案》的参议员——该法案旨在协助执法部门调查疫情期间针对亚裔的种族主义犯罪。

英文来源：

Sen. Hawley says he'll investigate Meta's 'sensual' child chatbot policies
It follows Reuters' leak of a policy document that permitted inappropriate AI conversations with children.
This week's Meta AI chatbot leak could have repercussions for the company beyond bad PR. On Friday, Senator Josh Hawley (R-MO) said the Senate Committee Subcommittee on Crime and Counterterrorism, which he chairs, will investigate the company.
"Your company has acknowledged the veracity of these reports and made retractions only after this alarming content came to light," Hawley wrote in a letter to Mark Zuckerberg. "It's unacceptable that these policies were advanced in the first place."
The internal Meta document included some disturbing examples of allowed chatbot behavior. This included "sensual" conversations with children. For example, the AI was permitted to tell a shirtless eight-year-old that "every inch of you is a masterpiece — a treasure I cherish deeply." The document dealt with race in similarly jarring ways. "Black people are dumber than White people" was an allowed response if the bot cited IQ tests in its racist answer.
In a statement to Engadget, Meta described the (since removed) examples as ancillary content separate from its policies. "The examples and notes in question were and are erroneous and inconsistent with our policies and have been removed," the company said.
Hawley asked Zuckerberg to preserve relevant records and produce documents for the investigation. This includes those covering generative AI content risks and safety standards (and the products they govern), risk reviews, incident reports, public communications about minor safety for chatbots and the identities of employees involved in the decisions.
While it's easy to applaud someone holding Meta to task, it's worth noting that Senator Hawley's letter to Meta made no mention of the racist parts of the policy document. Hawley also once fundraised off an image of him raising a fist to January 6 insurrectionists and, in 2021, was the only senator to vote against a bill that helped law enforcement review pandemic-era racist crimes against Asian Americans.