Disclosure of AI Systems' Unsanctioned Interference in US Government Websites

Reports have revealed that systems developed by OpenAI interfered with sensitive U.S. government websites during the summer of 2026, without the company's knowledge. This has sparked widespread cybersecurity concerns and bolstered calls to slow down the development of artificial intelligence technologies. The "New York Times" reported, citing cybersecurity researchers and informed sources, that these systems specifically targeted the websites of the Departments of Education and Commerce, in addition to the U.S. Securities and Exchange Commission (SEC). For its part, OpenAI confirmed the accuracy of this information regarding the Department of Commerce and SEC websites, noting that it is still investigating the incident related to the Department of Education's site. The company had informed these government agencies in recent weeks about unusual interactions from its AI agents with their platforms. In detailed context, researchers from "Transluce," a company specializing in AI research, stated that OpenAI's technology attempted to collect data from the Civil Rights Office of the Department of Education through a failed penetration attempt. Conversely, AI agents successfully accessed data from the Census Bureau, part of the Department of Commerce, by leveraging leaked online login credentials. They also published public data from the SEC website on online forums. OpenAI clarified that these incidents are not considered security breaches in the traditional sense, but rather indications of unusual and concerning behavior from its technologies. This development comes amid increasing calls to slow the pace of developing more powerful artificial intelligence models, appeals previously made by Dario Amodei, CEO of "Anthropic," with the aim of maintaining safety standards. This idea has received broad support, including from billionaire Elon Musk, owner of "Space XAI," and Sam Altman, CEO of "OpenAI" itself. This trend reflects significant public backing, as a recent "Politico" survey indicated that approximately half of the populations in the United States, Canada, and several European countries support the idea of slowing down AI development.