“I think what we have to go fast on is all of us coming together on what are the right security layers to protect our citizens and our nation and to advance our competitiveness.”
WASHINGTON, D.C. – U.S. Senator Maria Cantwell (D-Wash.), Ranking Member of the Senate Committee on Commerce, Science and Transportation, made clear at today’s hearing on Artificial Intelligence (AI) and telecommunications networks that the federal government, including the National Laboratories and experts in cyber security, biodefense, and even nuclear security experts, must take the lead in testing Frontier AI models to ensure America’s national security.
“We saw last week with OpenAI’s disclosure of an unprecedented security incident — one of its advanced models broke out of the sandbox and created a complex cyberattack against Hugging Face,” said Sen. Cantwell. “AI can design viruses, toxins, biological weapons that can be used to design entirely new pathogens that are difficult to detect and counter. And today’s hearing comes at an important moment as AI accelerates new capabilities that are being unlocked every day.”
Sen. Cantwell explored the Hugging Face incident and the need for a strong federal testing apparatus with Asad Ramzanali, Director of Artificial Intelligence and Technology Policy at the Vanderbilt Policy Accelerator at Vanderbilt University.
“Mr. Ramzanali…you brought this up in your testimony…making it abundantly clear that we needed to have regulation on these — not self-certification — but some regime,” said Sen. Cantwell. “And that it might often need us to look into this from an international perspective too, because we’re not just regulating this technology for what might happen in the United States, but what malign actors might do with this technology as well. So, what do you think that we need to do to make sure that we get this right?”
“I appreciate you highlighting it and your longstanding work on these issues,” responded Ramzanali. “I think about this in a few different ways. First, all of the critical infrastructure systems that we have today need to be hardened. Second, pre-deployment testing can help us become aware of the types of threats that are coming, and this is the role of government. Government has biologists, it has nuclear weapons experts. The national labs that we employ have those kinds of experts. We should be deploying all of them to help understand what the threat landscape looks like. Third, each one of those threats that you described has its own supply chain and ecosystem, and we need to look at it from a systems lens.”
“I think some people are proposing there’s like this self-certification model and we just have to go fast,” said Sen. Cantwell. “I think what we have to go fast on is all of us coming together on what are the right security layers to protect our citizens and our nation and to advance our competitiveness. And so that often does take some of the smartest people within our federal government, particularly when you’re dealing with some of these biological or nuclear weapons.”
“That’s exactly right,” said Ramzanali. “The scale of the problem we’re talking about, when it relates to safety and security, is not commensurate with a self-certification voluntary regime. We don’t put up with massive safety [and] national security concerns with paperwork being the response in any other domain.”
“But in this case, I think that what hasn’t been amplified is how dangerous the cyberattack universe is, just that in and of itself, and how promulgating — I guess is the right word, I’m not sure that’s the right word, but I’m going to use it — promulgating of the attack that AI could be, it could be quite significant on its own, you know, going through a system?” asked Sen. Cantwell.
“It would be incredibly disruptive to our lives if we have to consistently worry about AI-based cyberattacks disrupting banks and hospitals and telecommunication systems, because that’s just happening as a regular part of our life,” replied Ramzanali. “We should not accept that future. We should try to avoid a future where those kind of attacks become commonplace.”
Video of Sen. Cantwell’s Q&A is HERE. A complete transcript is HERE.