据《华盛顿邮报》报道，一家网络安全初创公司在不到12小时内获取了人工智能公司Anthropic的敏感内部数据。
A cybersecurity startup gained access to sensitive internal data at artificial intelligence company Anthropic in less than 12 hours, according to a report by The Washington Post .
据该报审阅的一份草案报告显示，由美国政府效率部（DOGE）前成员创立的数月大初创公司Cathedral，发现了Anthropic所使用的第三方商业软件中的漏洞。Anthropic是Claude聊天机器人的开发商。
Cathedral, a months-old startup founded by former members of the US Department of Government Efficiency (DOGE), identified vulnerabilities in third-party business software used by Anthropic, the maker of the Claude chatbot, according to a draft report reviewed by the newspaper.
报告称，研究人员利用AI代理寻找进入Anthropic内部系统的途径，获取了会议记录、薪资和工作录用详情，以及讨论先进AI模型的Slack消息。
The researchers used AI agents to find pathways into Anthropic’s internal systems, gaining access to meeting transcripts, salary and job offer details, and Slack messages discussing advanced AI models, the report said.
该测试是在允许外部网络安全研究人员探测Anthropic系统漏洞的计划下进行的。
The testing was conducted under a program that allows outside cybersecurity researchers to probe Anthropic’s systems for vulnerabilities.
Anthropic证实，Cathedral访问了其销售、人力资源和IT系统的“有限集合”。
Anthropic confirmed that Cathedral accessed a “limited set” of its sales, human resources and IT systems.
该公司表示，已关闭暴露的途径，并就该事件向联邦官员进行了通报。
The company said it had closed the exposed pathways and briefed federal officials about the incident.
Anthropic发言人称，Cathedral在转向第三方商业工具之前，曾花费数天试图入侵该公司自身的软件。
thropic spokesperson said Cathedral had spent several days attempting to breach the company’s own software before turning to third-party business tools.
一名特朗普政府高级官员据报道称，该报告已在政府内部流传，并引发了人们对该初创公司访问敏感系统速度的担忧。
A senior Trump administration official reportedly said the report had circulated within the government and prompted concern over the speed with which the startup accessed sensitive systems.
该官员警告称，这些发现引发了关于外国情报机构是否可能利用类似弱点的疑问。
The official warned that the findings raised questions about whether foreign intelligence agencies could exploit similar weaknesses.