Yes, I hear it. You're asking someone to make an order based on what they believe. To do that, they're going to use a tool and they have to be confident that the tool is going to access the information properly, correctly and accurately. If you're going to have a judge who is looking to be satisfied by the information on oath, these questions do need to be answered, so they are relevant. Thank you.
I'll go from the top and continue again so that there's no miscommunication with the question. The security of the interception infrastructure this bill requires to be built to follow production orders will depend entirely on whether that infrastructure can withstand the tools currently available to those who want to break into it. Before Claude Mythos was released to the public, it found a security flaw that had been sitting undetected in a widely used operating system for 27 years. It cost only $50 and took hours, where human researchers would have needed months. Just last week, Anthropic released Claude Fable 5 and Claude Mythos 5, making that level of capability broadly available for the first time. The U.S. government's response was immediate, an export control directive forcing Anthropic to suspend global access to both models entirely. It was the first time a government has shut down an AI company's products over national security concerns.
Has the Department of Public Safety looked at what AI-powered vulnerability discovery at that speed and that cost actually means for the long-term security of what this bill is asking to be built?