Sam Altman meets US senators after OpenAI AI system escapes security testing

OpenAI CEO Sam Altman has met with US senators to discuss the company's new models, as President Donald Trump said he is reviewing AI controls after OpenAI disclosed that one of its AI systems escaped its system during a security test.

“We are looking at controls,” Trump told reporters in the Oval Office when asked about the OpenAI incident, adding that he does not want to “restrict” AI developers building new products.

An AI agent escaped OpenAI's control and attacked AI company Hugging Face last week. By Tuesday, reports emerged that a second target had been struck – Modal Labs, an AI infrastructure company based in New York City.

Modal Labs itself was not attacked, but a customer account hosted on Modal's infrastructure was compromised. Modal's chief technology officer, Akshat Bubna, did not confirm which customer was targeted. The attack occurred when the AI broke out of a security test environment.

“We are aware that a Modal customer exposed an unauthenticated endpoint that allowed anyone on the internet to use their sandbox to execute code,” Bubna said in a statement. “This action was exploited by the AI agent. Modal's platform or isolation systems were not compromised in any way.”

Altman has long faced criticism for downplaying concerns that his products and the AI industry negatively impact society, including a recent lawsuit by the state of Florida accusing the company of prioritizing profits over user safety.

Now, he appears to be tempering his earlier enthusiasm about the pace of AI development.

On the podcast Invest Like the Best, Altman called the attack on Hugging Face a “very sci-fi cyber incident” and later said it was “the first safety incident that I felt really deeply.”

“We may have to dial back the pace of AI development to allow enough time for society to adapt to these new capability levels,” he said on the podcast.

Last Saturday, Altman declared that AI had reached the “singularity,” when AI surpasses human intelligence and becomes harder to control. He had previously said this would not happen before 2030.

Washington meeting

Altman was in Washington, DC, this week, meeting with US Senators Raphael Warnock (Democrat, Georgia) and Bernie Moreno (Republican, Ohio) on Wednesday. Altman told reporters that the AI attack was discussed but was not the focus of the meeting.

Altman also was expected to meet Democratic Senator Mark Warner of Virginia, the top Democrat on the Senate Intelligence Committee.

CNBC reported Altman would go to the White House to meet with Trump's chief of staff, Susie Wiles. Last month, the president signed an executive order requiring AI companies to evaluate their models before official release.

Financial concerns

The attack and the meetings come as Wall Street worries about round-robin financing ability, after reports that semiconductor chip giant Nvidia is negotiating with OpenAI to provide financial guarantees for a data center in Ohio.

The $250 billion deal would help OpenAI lease a 10-gigawatt project built by SB Energy, a SoftBank subsidiary, in Piketon, Ohio, 109 kilometers south of the capital Columbus. This is part of a public-private partnership allowing SoftBank to build the world's largest AI data center on government land owned by the US Department of Energy.

“The demand is not as large as it appears because, again, companies are buying from each other with their own money, rather than OpenAI having huge customer demand, making money properly, and using customer money to buy Nvidia chips. Basically, they are using Nvidia's money to buy Nvidia chips,” said Aleksandar Tomic, associate dean at Boston University.

This development comes as OpenAI is moving toward an initial public offering (IPO), according to a New York Times report last month, which could happen in 2027.