Autonomous hacks to US government websites lead OpenAI to suspend the training of its Artificial Intelligence agents

The decision to halt development came just hours after the company revealed on Friday that it was reviewing several recent incidents in which OpenAI agentswhen searching United States federal government websitesacted in unexpected ways beyond what was asked of them while collecting and distributing information.

Separately, AI tester Transluce indicated that agents appearing to come from OpenAI unsuccessfully attempted hack a Department of Education websitea detail that OpenAI has not confirmed.

OpenAI said in a statement that it will resume training when it has “the confidence to have additional safeguards”, adding that he anticipates he will have to “pause again” as AI advances and other problems arise.

AI labs face pressure from lawmakers and tech experts to slow development so they can build protective barriers that prevent agents from acting on their own, hacking websites, and ddisclose non-public information. The directors of both OpenAI, whose CEO is Sam Altman, and its rival Anthropic have also asked to go slower.

It is the second time in three months that OpenAI has stopped the development of its models. The first occurred in July, after the disclosure of a cyberattack targeted AI startup Hugging Facean incident that stoked fears that the industry was losing control.

In a meeting with the Chinese president Xi Jinping this week, the American president Donald Trump agreed share information about the dangers of AI and coordinate efforts to keep it safe. However, Trump considers fears about AI overblown and later hinted that he is not planning any offensive of his own.

The United States is not going to “hit the brakes,” Trump told reporters outside the White House. “They want to slow down our progress because we are far ahead of China, and we are going to keep it that way.”

OpenAI’s most recent incidents did not appear to involve the disclosure of non-public information, but were concerning enough for the company to warn the federal agencies involved.

In the Department of Education incident, OpenAI agents found API “developer keys” to access government data, although in the end only publicly available information was collected.

In another case involving to the Securities and Exchange Commission (SEC), The agents found information available to everyone, but then they published it somewhere else of the internet, an act that went beyond what had been indicated to them.

SEC spokesman Kurt Hopfenspirger stated this Saturday that “no non-public information was accessed.” The Department of Education previously said it found “no evidence of any impact on our website or databases.”

Several other AI companies have disclosed incidents in which their models went out of control and even websites were hacked. The case in the United States occurs after the complaint by the Australian government.

OpenAI CEO Sam Altman said in a social media post on Friday that the Hugging Face incident “remains the most serious event we have seen.”

OpenAI previously shared six other reports of “unexpected or concerning” behavior in AI models and introduced a framework for tracking, examining and disclosing cases.