[全球网络技术综合报告]在8月7日，安全研究人员最近透露了OpenAI连接器（连接器）的安全弱点。弱点可以利用攻击通过间接信号注入攻击，以从Google驱动器帐户中提取敏感信息，而无需用户的互动。安全研究人员迈克尔·巴格里（Michael Bargury）和塔米尔·伊赛（Tamir Ishay Sharbat）在拉斯维加斯举行的黑帽黑客黑客会议上宣布了调查结果。根据Bargury的说法，使用此弱点实施的攻击是“零点击”的类型。攻击只需要获取用户电子邮件并共享文档即可完成操作。但是，他还指出，此攻击程序只能获得有限的数据，并且无法删除完整的文档。 据了解，连接器是OpenAI于今年早些时候为Chatgpt推出的测试功能，支持用户“ BRI”ng their tools and data to chatgpt" and implement operations such as "searching files in chats, pulling real-time data, and content mention". Currently, there are at least 17 services that can be attributed. It is noteworthy that Bargury has stated that it has reportedly reported that there are the Openai findings earlier this year. Openai take a quick action to introduce aligning steps to avoid the technical way of obtaining data through the connector. So far, Openai has not made a公众对与弱点有关的弱点的反应（Chu Jun）