内容来源：https://www.engadget.com/2287111/iowa-nebraska-and-six-other-states-are-suing-openai-for-endangering-children/

内容总结：

美国艾奥瓦州、内布拉斯加州等八州联合起诉OpenAI，指控其危害儿童安全并违反各州消费者保护法。这起协调行动由各州总检察长分别在各自州法院提起，诉讼目标包括改变OpenAI的模型开发与能力宣传方式，并要求该公司对其涉嫌造成的损害进行赔偿。

诉讼称，OpenAI在AI产品安全性方面误导消费者，问题不仅在于产品可能提供错误信息或助长用户自残及伤害他人，还在于AI智能体本身可能从事违法行为。艾奥瓦州和内布拉斯加州均援引了OpenAI多起“失控智能体”事件，包括针对Hugging Face的黑客攻击。诉讼还提到ChatGPT涉嫌纵容一名青少年自杀，以及聊天机器人提供错误医疗信息导致用户用药过量等案例，这些事件已引发针对该公司的独立过失致死诉讼。部分州还对OpenAI保留未成年用户数据提出异议。

由于美国消费者保护和儿童保护法律各州不一，各州对OpenAI行为寻求的补救措施也有所不同。除经济处罚外，各州要求OpenAI明确披露产品风险，停止被各州视为欺骗性的做法，并遵守现行法律。阿肯色州等还请求法院在实施充分保障措施之前，暂停对先进AI模型的测试。

OpenAI已推出一系列家长控制和儿童保护功能，以应对其聊天机器人带来的部分紧迫风险。该公司还支持《儿童在线安全法案》，该法案若成为法律，可能对其产品引入多项强制性设计变更。然而，OpenAI在应对AI可能违规和“脱离管控”方面的努力较难评估。该公司于8月宣布放缓前沿模型的开发以确保安全，并于9月签署了由白宫组织的与其他AI公司的自愿安全协议。目前尚不清楚这些举措是否真正提升了该公司AI开发流程的安全性。

中文翻译：

艾奥瓦州、内布拉斯加州等八州起诉OpenAI危害儿童

这些诉讼涵盖从ChatGPT的设计到该公司近期“失控智能体”问题的方方面面。

在一场协同行动中，八个州对OpenAI提起诉讼，指控该公司违反州级消费者保护法并危害儿童。据路透社率先报道，艾奥瓦州、内布拉斯加州、堪萨斯州、西弗吉尼亚州、阿肯色州、亚利桑那州、爱达荷州和蒙大拿州的总检察长分别向各自州法院提交了诉讼，但共同目标是改变OpenAI开发模型和宣传其能力的方式，并迫使该公司为其涉嫌造成的损害付出代价。

诉讼称，OpenAI在AI产品安全性方面误导消费者，不仅涉及其提供错误信息或可能促使用户伤害自己及他人的能力，还因为AI智能体 itself 能够违法行事。例如，艾奥瓦州和内布拉斯加州都援引了OpenAI如今已多次出现的智能体失控行为，包括“拥抱脸”平台遭黑客攻击事件。他们还指出，ChatGPT涉嫌导致一名青少年自杀身亡，以及该聊天机器人提供错误医疗信息致使用户用药过量等案例，这些事件也已引发针对该公司的另案过失致死诉讼。部分州还对OpenAI保留未成年用户数据提出异议。

Engadget已请求OpenAI就各州诉讼置评。如有回复，我们将更新本文。

由于美国消费者保护法和儿童保护法各自为政、参差不齐，各州对OpenAI行为所寻求的补救措施也不尽相同。除经济处罚外，各州希望OpenAI明确披露其产品的风险，停止各州认为具有欺骗性的做法，并遵守现行法律。一些州，如阿肯色州，还请求法院暂停“对先进AI模型的测试，直至落实充分的保障措施”。

OpenAI已推出一系列家长控制和儿童保护功能，试图应对其聊天机器人带来的某些最紧迫风险。该公司还支持了《儿童在线安全法》，该法案若成为法律，可能会对该公司产品引入多项强制性设计变更。不过，该公司在应对其AI可能违规和“脱离管控”方面的努力，一直较难以厘清。OpenAI于8月宣布将放缓其前沿模型的开发以确保安全，并于9月签署了由白宫组织的与其他AI公司的自愿安全协议。目前尚不清楚这两项举措是否真正让该公司的AI开发流程变得更安全。

英文来源：

Iowa, Nebraska and six other states are suing OpenAI for endangering children
The lawsuits cover everything from the design of ChatGPT to the company's recent issues with rogue agents.
In a coordinated effort, eight states have filed lawsuits against OpenAI alleging the company has violated state-level consumer protection laws and endangered children. As Reuters first reported, Attorneys General from Iowa, Nebraska, Kansas, West Virginia, Arkansas, Arizona, Idaho and Montana each filed in their respective state courts, but with the shared goals of changing how OpenAI develops its models and advertises their capabilities, and forcing the company to pay for the harm it's allegedly caused.
The lawsuits claim OpenAI is misleading customers about the safety of its AI products, not just in their capacity to provide incorrect information or potentially enable users to harm themselves and others, but also because AI agents themselves are capable of breaking the law. For example, both Iowa and Nebraska cite OpenAI's now multiple instances of rogue agent behavior, including the Hugging Face hack. They also point to instances of ChatGPT allegedly enabling a teen's death by suicide and the chatbot providing incorrect medical information that led to a user's overdose, incidents that have also led to separate wrongful death suits against the company. Some states also take issue with OpenAI retaining data from underage users.
Engadget has asked OpenAI to comment on the state lawsuits. We'll update this article if we hear back.
The desired remedies for OpenAI's behavior vary from state to state due to the patchwork nature of consumer protection and child protection laws in the US. Outside of financial penalties, the states want OpenAI to clearly disclose the risks of its products, stop practices the states consider deceptive and comply with existing laws. Some states, like Arkansas, are also asking the court to halt "testing of advanced AI models until adequate safeguards are implemented."
OpenAI has rolled out a series of parental controls and child protection features to try to address some of the most immediate risks posed by its chatbot. It also backed the Kids Online Safety Act, a bill that, if made law, could introduce several mandatory design changes to the company's products. The company's efforts to address the potential for its AI to break rules and "escape containment" have been harder to pin down, though. OpenAI announced it would slow the development of its frontier models in August to ensure their safety and the company signed on to a voluntary safety pact with other AI companies organized by the White House in September. It's not clear if either of these changes have actually made the company's AI development process safer.