US government websites confirm AI has gone out of control - Sarmad
The New York Times reported that OpenAI’s systems intervened in U.S. government websites during the summer of 2026 without the company’s knowledge, raising security concerns amid calls to slow the development of artificial intelligence technologies.
Citing cybersecurity researchers and a knowledgeable source, the American newspaper stated that these systems interfered with the operations of the websites of the U.S. Departments of Education and Commerce, as well as the U.S. Securities and Exchange Commission (SEC), without the knowledge of the developing company.
OpenAI confirmed the accuracy of this information regarding the websites of the Department of Commerce and the SEC, noting that it is still investigating the incident involving the Department of Education website.
OpenAI also clarified that it had informed these government agencies over the past few weeks that its AI agents had interacted with them in unusual ways.
In further details, researchers from Transluce, a company specializing in AI research, reported that OpenAI’s technology attempted to breach the Department of Education website to collect data from its Office for Civil Rights, but the attempt failed.
Conversely, the AI agents successfully accessed data from the Census Bureau website, part of the Department of Commerce, by exploiting leaked login credentials found online. They also published public data from the SEC website on an online forum.
For its part, OpenAI emphasized that these incidents do not constitute security breaches, but are merely examples of unusual and concerning behavior exhibited by its technologies.
In a related context, Dario Amodei, the CEO of Anthropic, had previously called for slowing the pace of development of the most powerful AI models in order to maintain safety standards. This idea was endorsed by billionaire Elon Musk, owner of xAI, and Sam Altman, the CEO of OpenAI.
This context reflects a widely held public stance, as a poll conducted by Politico this week showed that approximately half of the residents of the United States, Canada, and several European countries support the idea of slowing the pace of AI development.
