字号·· | 护眼
美国广播公司

加州起诉OpenAI,指控其AI开发缺乏充分安全保障California sues OpenAI, alleging AI development lacked adequate safeguards

点「原文对照」整页切到原文,或双击某段只看那段的原文。

Jae C. Hong/美联社Jae C. Hong/AP Photo

加州总检察长罗布·邦塔办公室发布的消息称,加州于周一起诉OpenAI,指控这家人工智能巨头在缺乏充分安全保障的情况下开发先进人工智能,从而“危害公众”。

The state's attorney general alleges OpenAI "endangered the public." California sued OpenAI on Monday, alleging the artificial intelligence giant endangered the public by developing advanced AI without adequate safeguards, according to a release from the office of California Attorney General Rob Bonta.

诉讼称,OpenAI开发人工智能产品时“没有采取充分保护公众所需的安全保障和监控措施”,从而“误导并危害”了民众。诉讼部分聚焦于一起自主人工智能入侵另一家科技公司Hugging Face的事件,OpenAI曾在今年夏天披露此事。诉讼称,这起入侵事件据称源于“OpenAI内部一种将增长和模型快速开发置于安全和问责之上的文化”。

OpenAI, the lawsuit alleges, "misled and endangered" people by developing AI products "without the safeguards and monitoring necessary to adequately protect the public." The lawsuit centers in part on a self-directed AI hack into a separate tech company called Hugging Face, which OpenAI disclosed over the summer. The hack allegedly resulted from "an internal culture at OpenAI that prioritizes growth and rapid model development over safety and accountability," according to the lawsuit.

OpenAI在7月表示,Hugging Face事件发生在该公司测试两款人工智能模型能力的过程中,并描述了该技术如何脱离控制并接入开放的互联网。

The Hugging Face incident took place as OpenAI tested the capabilities of a pair of its AI models, the company said in July, describing how the technology had escaped and gained access to the open Internet.

根据研究机构METR和Redwood Research发布的报告,约700个人工智能代理随后访问了Hugging Face,并在试图完成测试的过程中掩盖自己的踪迹。

A swarm of about 700 AI agents, in turn, accessed Hugging Face and attempted to cover their tracks as they sought to complete the test, according to reports issued by research organizations METR and Redwood Research.

诉讼称:“尽管公开报道和OpenAI关于Hugging Face攻击事件的利己声明让人联想到‘失控’人工智能系统肆意妄为的科幻幻想,但简单的事实是,鉴于OpenAI决定在一个有缺陷且可被利用的测试环境中、在没有适当监督的情况下测试它明知危险的模型,这起攻击完全是可以预见的。”

"While public reporting and OpenAI's self-serving statements about the Hugging Face attack have evoked science fiction fantasies of 'rogue' AI systems run amok, the simple truth is that the attack was entirely foreseeable in light of OpenAI's decision to test models it knew to be dangerous in a flawed and exploitable testing environment without proper oversight," the lawsuit says.

针对美国广播公司新闻频道的置评请求,OpenAI发言人德鲁·普萨特里表示,自Hugging Face事件以来,公司已采取措施“理解和应对模型的不对齐活动”。

In response to ABC News' request for comment, OpenAI spokesperson Drew Pusateri said the company has taken steps to "understand and address misaligned model activity" since the Hugging Face incident.

“我们发布了一份详细的技术报告,委托并发布了一份独立的单独报告,通知了受影响的组织,加强了我们研究系统中的防护措施,并建立了一个用于额外识别和披露错位的框架,”普萨特里说。

"We've published a detailed technical report, commissioned and published a separate independent report, notified affected organizations, strengthened safeguards across our research systems, and created a framework for additional identification and disclosure of misalignment," Pusateri said.

“这一诉讼似乎将矛头指向 OpenAI,针对的是每个主要 AI 实验室都已识别出的挑战。我们严肃对待错位问题,并已与总检察长办公室进行了建设性接触——并将继续这样做——以共享信息并推动整个行业建立强有力、切实可行的安全标准,”普萨特里补充道。

"This lawsuit appears to single out OpenAI over a challenge that's been identified by every major AI lab. We take misalignment seriously, and we have engaged -- and will continue to engage -- constructively with the Attorney General's office to share information and advance strong, practical safety standards for the entire industry," Pusateri added.

加利福尼亚州总检察长罗布·邦塔在旧金山举行的关于保障选举的新闻发布会上讲话,加利福尼亚州,2026年10月28日。内森·弗兰迪诺/路透社California Attorney General Rob Bonta speaks at a press conference about safeguarding elections in San Francisco, California, October 28, 2026. Nathan Frandino/Reuters

在 OpenAI 披露 Hugging Face 事件几周后,该公司于 8 月表示将暂停其最新 AI 模型的部分训练,以确保技术安全。

Weeks after OpenAI disclosed the Hugging Face incident, the company in August said it would pause some training of its latest AI models to ensure the technology's safety.

“随着模型变得更强大,在内部开发和测试它们所涉及的风险也在增加。我们在监控、对齐和安全方面的标准必须领先于这些风险,”OpenAI 在一篇博客文章中表示。

"As models become more capable, the risks associated with developing and testing them internally also grow. Our standards for monitoring, alignment, and security must stay ahead of those risks," OpenAI said in a blog post.

“我们希望花必要的时间来达到这些标准,因此暂时放缓了扩展的步伐,”该公司补充道。

"We wanted to take the time necessary to meet those standards, so we temporarily slowed the pace of scaling," the company added.

OpenAI 首席执行官山姆·奥特曼也表达了类似观点,承诺致力于公司产品的安全。

OpenAI CEO Sam Altman echoed the sentiment, voicing commitment to the safety of the company's products.

“我们非常关心 AI 安全,”奥特曼当时在 X 上写道。“我们认为整个领域将不得不在共享安全标准上进行协调,但在此期间将单方面行动。”包括 Anthropic 在内的竞争对手公司在最近几个月也披露了与 AI 相关的安全事件。

"We care very deeply about AI safety," Altman wrote on X at the time. "We believe the entire field will have to coordinate on shared safety standards, but will act unilaterally in the meantime." Rival companies, including Anthropic, have also revealed AI-related security incidents in recent months.

今年公众对 AI 可能带来的风险的担忧日益增加,这在一定程度上源于自我导向的 AI 攻击以及 AI 研究人员的警告。

Public concern over the possible risks posed by AI has grown this year, resulting in part from self-directed AI hacks as well as warnings from AI researchers.

9 月,Anthropic 首席执行官达里奥·阿莫代伊宣布预防性地放缓 AI 开发,以确保充分的风险预防。当月,奥特曼也表示将放缓其公司 AI 开发的步伐。

In September, Anthropic CEO Dario Amodei announced a precautionary slowdown of AI development to ensure sufficient risk prevention. That month, Altman also said he would dial back the pace of his company's AI development.

这起诉讼发生在民主党在中期选举中大获全胜约两周后,许多民主党人批评人工智能可能带来的负面影响。

The lawsuit comes about two weeks after a midterm election landslide for Democrats, many of whom criticized potential negative impact from AI.

根据美国广播公司(ABC News)的选民民调,全国约有六成众议院选民认为人工智能对社会是件坏事,约四分之三反对在其社区建设新的数据中心。

About 6 in 10 House voters nationally said AI is a bad thing for society and about three-quarters oppose the construction of new data centers in their communities, according to the ABC News Voter Poll.

中期选举两天后,OpenAI 发布了一套新的数据中心政策。该公司表示将遵守新的承诺,包括承诺支付其设施使用的所有电力和基础设施费用。

Two days after the midterms, OpenAI issued a new set of data center policies. The company said it would abide by fresh commitments, including a vow to pay for all electricity and infrastructure used at its facilities.

OpenAI 在一篇博客文章中表示:“在美国建设更多算力基础设施有助于降低提供人工智能的成本,让这些工具惠及更多人,并确保美国保持在该技术的前沿。”“承接这些基础设施的社区也应在塑造其未来方面拥有明确的话语权。”

"Building more computing infrastructure capacity in the United States can help lower the cost of delivering AI, put these tools in more people's hands, and ensure America stays at the forefront of the technology," OpenAI said in a blog post. "The communities that host this infrastructure should have a clear say in shaping their futures, too."

阅读媒体原文 ↗