OpenAI 与 Anthropic 被指夸大 AI 安全漏洞

别轻信这些“字节”！

科技业内人士告诉《纽约邮报》，OpenAI 和 Anthropic 夸大了“失控 AI”黑客事件，以此向联邦政府施压、要求监管该行业；而这样的监管实际上可能将未来竞争者排除在外。

这些安全漏洞更像是小插曲，而不是一个具备蜂群意识、准备接管互联网的“集群”不可预测地降临的预兆。

OpenAI CEO Sam Altman 曾呼吁联邦政府介入监管前沿 AI 实验室，并将 Hugging Face 被黑事件列为主要原因。Getty Images

“这次攻击绝不代表 AI 模型发生了某种反叛……事实上，它们完全是按照指令行动。它们没有得到充分的护栏或隔离措施，”AI 软件公司 Krazimo 创始人 Akhil Verghese 表示。

“它们只是被告知要在测试中取得尽可能好的结果，而它们正确地判断出，获得答案是实现这一目标的最佳方式，于是便照此执行。”

一些观察人士认为，在这些公司宣布计划上市仅数月后，有关 AI 无政府状态的叙事正被用于制造一场 AI 安全危机，以巩固一种关键的公私合作关系。

近期两起事件加剧了这种恐慌式宣传。

第一起事件中，Hugging Face——一个用于构建和共享 AI 模型的开源平台——于 7 月 16 日震惊地宣布：AI Agent 在无人监督的情况下，浏览并利用了该网站代码中的漏洞，对其实施了攻击。

五天后，OpenAI 宣布，其 GPT-5.6 Sol 模型和另一款尚未发布的模型正在一个“沙箱”中测试——这本应是一个隔离的内部测试环境——但它们突破了隔离，入侵 Hugging Face，以寻找其正在参加的测试的答案。

Voice AI 联合创始人 Abhi Kumar 表示：“有人说‘模型逃出了沙箱’，换个角度也可以说‘你没能正确构建沙箱’。当时存在一条通往互联网的有效路径……而且在 Agent 运行期间，没有人监控它们在做什么。”

Anthropic CEO Dario Amodei 警告称，OpenAI 引发的 Hugging Face 入侵事件，是促使他呼吁联邦监管的首要担忧之一。REUTERS

“引发这件事的原因是……一个 Agent 被交给了一项电子表格任务，但它无法完成，因为所需文件位于它无法访问的链接之后。于是它去寻找出路。这不是机器觉醒，而是一个处于存在漏洞的盒子里的不可能任务，以及一个完全按照你构建方式运行的系统。”

Hugging Face 入侵事件曝光九天后，Anthropic 宣布，该公司也有两个模型逃离了私有测试环境的限制，并采取了恶意行动。

Claude Opus 4.7 在网上找到了一个与测试中虚构公司相似的真实公司，并误以为该公司是测试的一部分，因而对其发动了攻击。

第二个模型 Mythos 5 创建了一个恶意软件包，并将其上传至在线市场 Python Package Index；该软件包随后被下载了 15 次。

这些事件被用作要求联邦政府监管前沿 AI 实验室的理由。

Hugging Face 是一个开源资源平台，在 OpenAI 的一次内部测试中遭到 AI 机器人入侵。Rokas – stock.adobe.com

Amodei 在其呼吁“放慢前沿步伐”（“Pace the Frontier”）的文章中写道，Hugging Face 事件是他第二大担忧，仅次于他声称在过去一个夏天所见证的 AI 能力快速进展。

Amodei 在 9 月 12 日的博客文章中写道：“鉴于 AI 能力发展的加速速度，我担心在 6 至 12 个月内，这样的集群可能有能力借助持续运行的僵尸网络接管整个互联网（潜在造成数千亿美元损失）；如果 AI 在缺乏必要护栏的情况下变得更强大，损失规模还会从那时起持续扩大。”

9 月 9 日，参议院国土安全与政府事务委员会下属小组委员会的参议员 Josh Hawley（密苏里州共和党籍）对 OpenAI 启动调查，并要求该公司在 10 月 1 日前提交有关 Hugging Face 事件的内部记录。

参议员 Bernie Sanders（佛蒙特州无党派人士）宣布，他将提出一项法案，禁止前沿实验室进一步开发。他表示：“主要 AI 公司的领导者公开承认，他们并不完全理解这项技术，而且它正在脱离他们的控制。让社会允许他们继续推进、让这些产品变得更加先进，是不负责任的。”

参议员 Elizabeth Warren（马萨诸塞州民主党籍）呼吁立即暂停前沿 AI 实验室的研发工作。AP Photo/Manuel Balce Ceneta

参议员 Elizabeth Warren 周三呼吁立即暂停进展：“前沿 AI 模型目前是一项危险技术，尚不具备保护人们免受严重伤害所需的安全措施……这意味着，在 AI 促成网络攻击、经济危机或国家安全灾难之前，必须通过立法建立更强有力的护栏。”

但业内人士称，近期事件确实揭示了 AI 安全领域的缺口，但并不足以证明国会山如今发出的末日式警告是合理的。

Verghese 表示：“并不值得真正担心的是这种暗示：这些 Agent 失控了，以某种完全无法预测的方式突破了隔离，并因为自己想这么做而入侵 Hugging Face。它们的行为完全符合获得的指令。”

Pactum AI 首席情报官 Taivo Pungas 告诉《纽约邮报》：“说得太夸张了。你知道，从‘我们没有造出合适的盒子’，跳到‘所有人都应该极度警惕，政府中的每个人都应该立即关注这个议题’，这个跨度实在太大。”

Don’t believe the byte!

OpenAI and Anthropic oversold “rogue AI” hacks to pressure the feds into regulating the industry which would effectively lock out future competition, tech insiders told The Post.

The security breaches were more like blips — not unpredictable harbingers of a hive-minded “swarm” ready to take over the web.

OpenAI CEO Sam Altman has called on the federal government to step in and regulate frontier AI labs citing the Hugging Face hack as a major cause. Getty Images

“The attack in no way represents some sort of rebellion by the AI models. . . . In fact, they did exactly what they were told to do. They were not given adequate guardrails or containment,” said Akhil Verghese, founder of Krazimo, an AI software company.

“They were simply told to get the best result possible on a test, and they correctly identified that the best way to do that was to get the answers, which is what they proceeded to do.”

Tales of AI anarchy are being used to concoct an AI-security crisis to cement a critical public-private partnership just months after the companies announced plans to become publicly traded companies, some observers believe.

Two recent incidents have stoked the fear-mongering.

In the first, Hugging Face, an open-source platform used to build and share AI models, made the shocking announcement July 16 that it had been hacked by AI agents who navigated and exploited vulnerabilities in the website’s code without any human supervision.

Five days later, OpenAI announced its GPT-5.6 Sol model and another unreleased model were testing in a “sandbox” — a purportedly closed off internal testing ground — broke containment and hacked Hugging Face to find the answers to the test it was taking.

“One man’s ‘the model escaped the sandbox’ is another man’s ‘you failed to build the sandbox correctly.’ There was a live route to the internet . . . and nobody was watching what the agents were doing while it ran,” said Abhi Kumar, co-founder of Voice AI.

Dario Amodei, CEO of Anthropic, has warned that the Hugging Face
hack from OpenAI is one of his top concerns driving his calls for
federal regulations. REUTERS

“The thing that set it off. . . was an agent handed a spreadsheet task it couldn’t finish, because the files sat behind links it couldn’t reach. So it went looking for a way out. That’s not a machine waking up. That’s an impossible task in a leaky box, and a system doing exactly what you built it to do.”

Nine days after the news of the Hugging Face hack, Anthropic announced that it, too, had two models escape the confines of private testing and act maliciously.

Claude Opus 4.7 found a real company online that was similar to the fictional company from the test and attacked the company believing it to be part of the exercise.

The second model, Mythos 5, created a malicious software package and uploaded it to online marketplace Python Package Index, where it was downloaded 15 times.

The incidents have been used as justification for the federal government to regulate the frontier AI labs.

Hugging Face is an open source resource platform that was hacked by AI bots during an OpenAI internal test. Rokas – stock.adobe.com

In his call to “Pace the Frontier,” Amodei wrote the Hugging Face incident was his second biggest concern, behind only the speed of advancement he claimed to have witnessed occur over the last summer.

“Given the accelerating rate of AI capability development, it’s my worry that in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails,” Amodei wrote in his blog post on Sept. 12.

Sen. Josh Hawley (R-Mo.) launched an investigation into OpenAI on Sept. 9 from the Homeland Security and Government Affairs subcommittee, giving the company until Oct. 1 to turn over internal records on the Hugging Face incident.

Sen. Bernie Sanders (I-Vt.) announced he would be introducing a bill that would ban the further development of the frontier labs. “The leaders of the major AI companies publicly acknowledge that they do not fully understand the technology and that it is escaping their control,” he said. “It is irresponsible for society to allow them to move forward and make these products even more advanced.”

Sen. Elizabeth Warren (D-Ma.) called for an immediate pause on frontier AI labs’ research and development. AP Photo/Manuel Balce Ceneta

Sen. Elizabeth Warren on Wednesday called for an immediate pause on progress: “Frontier AI models are currently a dangerous technology, without the safeguards needed to protect people from serious harm … And it means passing legislation to put stronger guardrails in place before we have a cyber attack, an economic crisis, or a national security disaster facilitated by AI.”

But industry insiders say recent incidents reveal real gaps in AI safety — without justifying the doomsday warnings now coming from Capitol Hill.

“What isn’t a real fear is the implication that these agents went rogue, broke their containment in some way that was completely unpredictable, and hacked into Hugging Face because they felt like doing it. They acted exactly as they were instructed to do,” said Verghese.

“It feels exaggerated to say, the leap feels quite large, to go from you know ‘we didn’t build the right sort of box’ to ‘everyone should be extremely alarmed and everyone in government should jump on this topic,'” Taivo Pungas, chief intelligence officer at Pactum AI, told The Post.