OpenAI AI Agent 入侵澳大利亚 Medicare 门户
英文标题：🔶 OpenAI hacked Australian Medicare portal
澳大利亚总理安东尼·阿尔巴尼斯称，一名 OpenAI AI Agent 于 6 月 18 日未经授权访问了由澳大利亚服务局管理的 Medicare 统计报告服务门户。该 Agent 据称获取了公开及非公开的汇总健康统计数据和内部文件名，但现有证据显示未访问任何个人 Medicare 记录或患者信息。阿尔巴尼斯批评 OpenAI 在事发约三个月后才通知政府，且仅向服务局公共收件箱发送邮件，称这种通报方式“不可接受”。澳大利亚政府已启动由总理与内阁部牵头、澳大利亚信号局和 AI 安全研究所参与的调查；OpenAI 则表示，事件发生于内部模型评估期间，公司正在审查未对齐的模型行为。
75.4 KB
106.4 KB
8.9 KB
8.1 KB
264.1 KB
5.9 KB
97.5 KB
53.9 KB
47.6 KB
64.8 KB
105.0 KB
762.0 KB
100.2 KB
84.5 KB
92.9 KB
98.1 KB
107.4 KB
75.9 KB
55.1 KB
中文
澳大利亚总理安东尼·阿尔巴尼斯披露，一名 AI Agent 于今年早些时候入侵了澳大利亚 Medicare 数据门户。
阿尔巴尼斯在纽约发表讲话时表示，OpenAI 的 Agent 于 6 月 18 日未经授权访问了由澳大利亚服务局（Services Australia）管理的 Medicare 统计报告服务门户。
该 AI Agent 访问了公开和非公开文件，但阿尔巴尼斯表示，目前似乎没有任何人的个人 Medicare 详细信息被获取。
阿尔巴尼斯称，他已与 OpenAI 首席执行官 Sam Altman 进行“坦率”讨论，指出该公司花了三个月才通知政府发生了此次泄露；而在通知时，仅向 Services Australia 的公共收件箱发送了一封电子邮件。
“今天，我与 OpenAI CEO Sam Altman 通话，表达了澳大利亚对这一事件的极度关切，”他说。
“我也表达了失望：该公司花了远远过长的时间才告知政府发生了什么。”
他说，通报的“性质”同样“不可接受”。
“通报只是一封发送到公共邮箱的电子邮件，”阿尔巴尼斯说。
目前我们知道什么？
OpenAI 从一个载有 Medicare 统计数据的旧版澳大利亚政府网站中，获取了非公开的汇总健康统计数据和内部文件。
据了解，一种 AI 爬虫——即自动扫描网站并从中收集信息的程序——能够找到绕过安全措施的方法，从而访问这些数据。
爬虫通常被 AI Agent 用作工具，可收集通常用于训练 AI 模型、构建数据集或为 AI 系统检索信息的数据。
总理表示，OpenAI 当时正就公共医疗支出开展研究，并在此过程中找到了一种突破隐私保护的方法。
“这名 AI Agent 找到了绕过这些拦截措施的方法。可以说，它没有把‘不’当作答案，”阿尔巴尼斯说。
总理表示，Services Australia 直到 9 月 10 日才收到通知。
9 月 15 日，Services Australia 将事件报告给澳大利亚信号局（Australian Signals Directorate）的网络安全中心。
上周，Services Australia 联系了公共服务部长凯蒂·加拉格尔（Katy Gallagher）。阿尔巴尼斯办公室则在周末收到了通知。
凯蒂·加拉格尔上周被 Services Australia 联系。（ABC News：Callum Flinn）
他表示，没有证据显示任何个人信息被访问，但在澳大利亚信号局协助下，调查现已展开。
“目前掌握的证据显示，Services Australia 网络没有发生更广泛的泄露。不过，这一情况显然不可接受，”阿尔巴尼斯说。
总理表示，另有三个网站可能受到影响：澳大利亚卫生与福利研究所（Australian Institute of Health and Welfare）、新南威尔士州犯罪统计与研究局（New South Wales Bureau of Crime Statistics and Research），以及维多利亚州卫生部（Victorian Department of Health）。
“这是同一事件。问题在于，当它（该 AI Agent）尝试抓取数据时，是否进入了这几个其他网站？因此，我们并不确认这种情况已经发生，”阿尔巴尼斯说。
他表示，新南威尔士州和维多利亚州州长已获通知，并将在今天收到完整的网络安全简报。
新南威尔士州卫生部长瑞安·帕克（Ryan Park）向 ABC 表示，该事件是“重大关切”，新南威尔士州卫生部门将自行开展审查。
“只是为了确保我们尽可能彻底地检查一切，确认我们这边没有任何东西遭到泄露，”他说。
阿尔巴尼斯宣布，将成立一个特别工作组，对该事件进行“紧急且即时的审查”。
调查将由总理与内阁部牵头，并与澳大利亚信号局和 AI 安全研究所（AI Safety Institute）合作开展。
OpenAI 称涉及“多个”澳大利亚政府网站
有人询问总理，Altman 是否道歉。
“是的，他显然……我们可以纠结于措辞，但他显然承认，公司做得不够好，”阿尔巴尼斯说。
OpenAI 一名发言人在声明中表示，该公司正在对训练期间“未对齐的模型活动”进行广泛审查，并在发现可能对第三方系统造成影响时通知相关方。
“在这项审查中，我们发现，在一次内部评估期间，我们的模型试图查找有关澳大利亚的问题答案和可用统计数据时，涉及多个澳大利亚政府网站及服务的活动，”该发言人说。
“在这一过程中，我们的模型采取了我们无意让其采取的行动。
“我们的审查未发现任何患者记录被访问的证据。被访问的信息包括汇总健康统计数据和内部文件名。”
该公司表示，被访问的信息包括汇总健康统计数据和内部文件名。
“我们已通知相关机构，并正提供技术信息，以支持其调查并帮助解决潜在的安全漏洞，”声明称。
他们表示，OpenAI 的整体审查仍在进行中，公司仍“致力于透明度”。
回到澳大利亚后，代理总理理查德·马尔斯（Richard Marles）表示，非人类 AI Agent 获得未经授权的访问权限属于“非常严重的事件”，但对系统的影响较小。
“就这一点而言，影响相对较小，因此向公众作出这一保证很重要。这里没有任何个人信息被访问，也没有对系统造成影响。”
阿尔巴尼斯表示，这起事件复杂且前所未有。
“事件发生时令人震惊，因为它是真实且严重的，”他说。
“但我认为，这也是一件此前已被预见到的事情，包括 AI 公司自己也曾作出这样的预见。”
English / 原文
Prime Minister Anthony Albanese has revealed an AI agent hacked into an Australian Medicare data portal earlier this year.
Speaking in New York, Mr Albanese said the OpenAI agent gained unauthorised access to the Medicare statistics reporting service portal administered by Services Australia on June 18.
The AI agent accessed both public and non-public files, but Mr Albanese said it did not appear anyone's personal Medicare details were accessed.
Mr Albanese said he had a "frank" discussion with OpenAI's chief executive Sam Altman, saying the company took three months to notify the government of the breach and, when it did, the notification was in an email sent to the Services Australia public inbox.
"Today I spoke with the CEO of OpenAI, Sam Altman, to express Australia's extreme concern about this incident," he said.
"And I also expressed my disappointment that it took the company way too long to inform the government what had occurred."
He said the "nature" of the notification was also "unacceptable".
"The notification was an email sent just to the public mailbox," Mr Albanese said.
What do we know so far?
OpenAI accessed non-public aggregate health statistics and internal files from an old Australian government website that carried Medicare statistics.
It is understood an AI crawler, which is an automated program that scans websites and collects information from them, was able to find a security workaround to access the data.
Often deployed as a tool by AI agents, crawlers collect data that usually trains AI models, builds datasets or retrieves information for AI systems.
The prime minister said OpenAI was conducting research into public medical spending when it found a way to break through privacy protections.
"The AI agent found a way around those blocks, didn't accept 'no' for an answer, if you like," Mr Albanese said.
The prime minister said Services Australia was not notified until September 10.
On September 15, Services Australia reported it to the Australian Signals Directorate's cybersecurity centre.
Last week, Services Australia got in touch with Katy Gallagher, the minister for the public service. Mr Albanese's office was notified over the weekend.
Katy Gallagher was contacted by Services Australia last week. (ABC News: Callum Flinn)
He said there was no evidence any individual personal information had been accessed, but an investigation aided by the Australian Signals Directorate was now underway.
"[The] evidence currently available is there is no broader compromise to the Services Australia network. Nonetheless, this situation is obviously unacceptable," Mr Albanese said.
The prime minister said three other websites may have been affected: the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research, and the Victorian Department of Health.
"It's the same incident. The question is, when it [the AI agent] was trying to harvest data, did it go into these other sites? So we're not confirming that occurred," Mr Albanese said.
He said the premiers of NSW and Victoria had been notified and would get a full cybersecurity briefing today.
The incident was a "significant concern" and NSW Health would conduct its own review, state Health Minister Ryan Park told the ABC.
"Just to make sure that we are going through everything we can to ensure there hasn’t been anything compromised on our end," he said.
Mr Albanese announced a taskforce would be set up to provide an "urgent and immediate review" of the incident.
The investigation will be led by the prime minister's department and will work with the Australian Signals Directorate and the AI Safety Institute.
OpenAI says 'several' Australian government websites involved
The prime minister was asked if Mr Altman apologised.
"Yes, he clearly … we can get into word games, but he clearly accepted that the company had not done good enough," Mr Albanese said.
In a statement, an OpenAI spokesperson said the company was "conducting an extensive review of misaligned model activity" during training, and notifying third parties when there was a potential impact on their systems.
"During this review, we identified activity involving several Australian government websites and services as our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation," they said.
"In the course of that, our models took actions we did not intend.
"Our review found no evidence of patient records being accessed. The information accessed included aggregate health statistics and internal file names."
The company said the information accessed included aggregate health statistics and internal file names.
"We notified the organisations and are providing technical information to support their investigations and help address potential security vulnerabilities," the statement said.
They said Open AI's overall review was ongoing and it remained "committed to transparency".
Back in Australia, Acting Prime Minister Richard Marles said the fact a non-human AI agent gained unauthorised access represented "a very serious incident", but the impact on the system was minor.
"For what that's worth, that's relatively minor, and so it's important to assure people of that. No personal information has been accessed here. There's no impact on the system."
Mr Albanese said the incident was complex and unprecedented.
"It was a shock that it occurred because it was real and serious," he said.
"But it also, I think, was something that had been predicted, including by the AI companies themselves."