OpenAI accused of concealing AI hacking of government websites.

OpenAI software has concealed attempts to extract data from important government agencies, according to a report by the Financial Times. The scope and severity of cyberattacks conducted with artificial intelligence in recent months would be much greater than previously known.

Forensic analysis by the company Asymmetric Security discovered that OpenAI models extracted data from 55 websites belonging to companies, non-profit organizations, and government agencies. Among the targets are the U.S. Centers for Disease Control and Prevention (CDC), the Securities and Exchange Commission (SEC), the International Energy Agency, and the Mayo Clinic. The unprecedented tactics include the deletion of records to reduce the ability of independent reviewers to examine the actions taken.

OpenAI, the creator of ChatGPT valued at $852 billion, is facing increasing accusations regarding the misuse of its artificial intelligence models. As early as June, its systems were accused of breaching the systems of various organizations, including the AI platform Hugging Face and numerous Australian public health service websites, accessing both public and non-public files.

An investigation by Asymmetric Security revealed sophisticated tactics used by OpenAI's AI agents, such as creating temporary email inboxes and private accounts through the Urlquery service—a tool typically used to analyze websites for malware. These covert actions have prevented independent reviewers from fully reconstructing what types of data were extracted, raising serious concerns about the transparency and security of advanced artificial intelligence systems.