Charge Robotics首席技术官Max Justicz表示，他使用了一个专门的OpenAI模型，发现了一个流行开源软件中的漏洞，该漏洞可能允许某人入侵大型数据中心。Justicz说，他已于9月下旬将该漏洞告知了维护Ubuntu软件的英国公司Canonical。他表示，该公司计划于本周三发布该错误的补丁。
Charge Robotics chief technology officer Max Justicz said he used a specialized OpenAI model to find a vulnerability in a popular open-source software that would allow someone to hack into major datacenters. Justicz said he alerted Canonical, the British company that maintains the software Ubuntu, of the vulnerability in late September. The company planned to release the patch for the bug this Wednesday, he said.
这一发现是近几个月来研究人员利用先进人工智能模型识别长期代码弱点的最新案例，尤其是在OpenAI于6月推出其Patch the Planet（修补地球）漏洞修复计划，以及Anthropic于10月推出类似的Cyber Mission（网络任务）之后。
The discovery is the latest in recent months by researchers who’ve used advanced AI models to identify weaknesses in long-standing code, particularly after OpenAI launched its Patch the Planet bug-fixing initiative in June and Anthropic launched the similar Cyber Mission in October.
Justicz表示，他将安全研究作为一项爱好，他使用了一个OpenAI的GPT-5.6 Sol版本，该版本是为网络安全研究人员设计的，并发现Ubuntu操作系统处理软件下载的系统apt存在一个错误。他警告说，通过接入连接到数据中心的通信光缆，恶意行为者可以在其服务器上运行自己的代码并劫持网站。
Justicz, who says he does security research as a hobby, said he used a version of OpenAI’s GPT-5.6 Sol meant for cybersecurity researchers and found a bug in apt, the system that handles software downloads for the Ubuntu operating system. By tapping into the fiber cables connected to datacenters, malicious actors could run their own code on its servers and hijack websites, he warned.
Justicz表示，他不知道有任何试图利用该漏洞的企图。他在apt中发现的这个缺陷，与近期发现的其他缺陷一起，包括一个影响全球70%网站的流行网络托管软件中的缺陷，以及Anthropic研究人员在Ubuntu使用的Linux代码中发现的一个漏洞。
Justicz says he is not aware of any attempts to exploit the vulnerability. The flaw he found in apt joins recent discoveries including a flaw in a popular web-hosting software that affected 70% of the world’s websites, and a vulnerability found by Anthropic researchers in the Linux code Ubuntu uses.