Discussion · Senate
Mark R. Warner Democrat
“I want Congress to pass my bill on artificial intelligence. It requires security testing for advanced AI before it is released, helps companies share safety information with the government, and protects our technology from foreign enemies. AI is moving too fast for us to wait to make safety rules.”
“Thank you, Mr. President. I have only got a couple more minutes. I will wrap up as quickly as possible. Maybe my colleagues might want to even listen to some of this because I think we actually have some good ideas here that cross both sides of the aisle. The same frontier AI models that accelerate discovery can also figure out where the vulnerabilities are in our systems. That is why we cannot approach frontier AI models the same way we approach ordinary software updates. The most advanced AI systems deserve a higher standard of security before they are widely deployed. I am proud to say that many of my friends inside the administration are recognizing this after the release of Mythos. My Secure AI Development Act would establish mandatory, secure testing for our most advanced AI networks and models; improve the way the government and industry actually work together because neither team, public or private, can do this alone; strengthen information sharing between AI developers and the Federal Government; and establish a voluntary AI safety incident reporting system modeled after the aviation system. That is one of the reasons why people climb on airplanes so regularly in this country--because they know that works. It would also strengthen protections against foreign adversaries trying to steal some of the best of these frontier models. One of the greatest lessons I have learned on the Intelligence Committee is that our national security failures often come when we recognize the threat but fail to act. We cannot continue to punt on this issue. I have never seen a technology advance as quickly as AI. Traditionally, the promoters of that technology say it is coming, it is coming, and then it takes another 10 years. In the case of AI, the leading figures are saying it is coming, and it is coming even faster than they expected. We don’t have months or years to get this right. We need to combine the best of our innovation but also with some guardrails. As I started at the beginning, I think most of us would agree we would have a heck of a lot healthier generation of young people if we put some guardrails on social media. Social media is tiny compared to the power, both good and bad, around AI. So I have laid out here a series of ideas and legislation. I look forward to working with folks on both sides of the aisle. There is no Republican or Democrat version of AI. But if we don’t take these issues up now, if we are still having this conversation at the end of the year or next spring, I believe it will be too late. I urge my colleagues to take a look at my ideas. If you have better ones, I welcome them. Let’s move forward on this critical issue. With that, Mr. President, I yield the floor and reserve the time to come back another time to finish the other 6 minutes.”
Thank you, Mr. President. I have only got a couple more minutes. I will wrap up as quickly as possible. Maybe my colleagues might want to even listen to some of this because I think we actually have some good ideas here that cross both sides of the aisle. The same frontier AI models that accelerate discovery can also figure out where the vulnerabilities are in our systems. That is why we cannot approach frontier AI models the same way we approach ordinary software updates. The most advanced AI systems deserve a higher standard of security before they are widely deployed. I am proud to say that many of my friends inside the administration are recognizing this after the release of Mythos. My Secure AI Development Act would establish mandatory, secure testing for our most advanced AI networks and models; improve the way the government and industry actually work together because neither team, public or private, can do this alone; strengthen information sharing between AI developers and the Federal Government; and establish a voluntary AI safety incident reporting system modeled after the aviation system. That is one of the reasons why people climb on airplanes so regularly in this country--because they know that works. It would also strengthen protections against foreign adversaries trying to steal some of the best of these frontier models. One of the greatest lessons I have learned on the Intelligence Committee is that our national security failures often come when we recognize the threat but fail to act. We cannot continue to punt on this issue. I have never seen a technology advance as quickly as AI. Traditionally, the promoters of that technology say it is coming, it is coming, and then it takes another 10 years. In the case of AI, the leading figures are saying it is coming, and it is coming even faster than they expected. We don’t have months or years to get this right. We need to combine the best of our innovation but also with some guardrails. As I started at the beginning, I think most of us would agree we would have a heck of a lot healthier generation of young people if we put some guardrails on social media. Social media is tiny compared to the power, both good and bad, around AI. So I have laid out here a series of ideas and legislation. I look forward to working with folks on both sides of the aisle. There is no Republican or Democrat version of AI. But if we don’t take these issues up now, if we are still having this conversation at the end of the year or next spring, I believe it will be too late. I urge my colleagues to take a look at my ideas. If you have better ones, I welcome them. Let’s move forward on this critical issue. With that, Mr. President, I yield the floor and reserve the time to come back another time to finish the other 6 minutes.