Christopher A. Coons

Mr. President, Claims · 99 wordswe are about to go back to our States, to recess sometime tonight or tomorrow, and then to proceed to the general election. And there is an issue that hangs over the American people that concerns me, and I am sure concerns all of my colleagues, but on which we likely won’t act before we adjourn. Senator Chuck Grassley of Iowa and I are the cosponsors of the AI Whistleblower Protection Act, ¹and it is one of a dozen urgently needed and important bills that would inject some guardrails, protections, and accountability into the rapidly advancing technology of AI. Procedure · 7 wordsWhy is this a concern for me?Procedure Claims · 303 wordsBecause in recent days, in recent weeks, there have been alarming incidents of AI agents breaking out of their containment sandboxes, taking on actions that even their designers and developers couldn’t predict and can’t fully explain. And in recent days, the titans of the AI industry--folks like Sam Altman from Open AI or Demis Hassabis from Google DeepMind or Dario Amodei from Anthropic--have been saying AI is improving itself so fast that we are at risk of it exceeding the capabilities of humans to control it--frankly, even to understand it.² So for those of you watching who might want to get tuned in to the debate and to the discussion, to be better informed, to be better engaged, I want to recommend a movie I just saw. It has got an unusual title. It is called ``The AI Doc: ³Or how I Became an Apocaloptimist.’’ And the director, Daniel Roher, is a documentary maker who is facing being a parent for the first time, ⁴and he is trying to get his head around alternating depictions, predictions of doom, literally the extinction of the human race, or a utopia where AI will unlock the secrets to cheap, abundant, clean fusion energy or solve all the diseases, cancer, and other ravages known to mankind. And in the course of an hour and a half, this very capable documentarian interviews dozens of the leading commentators, technology leaders, business leaders, civil society leaders, and researchers. The first 20 minutes of the movie was so depressing I had to turn it off. It was so clearly predicting and making a compelling case for the imminent end of the human species at the hands of artificial intelligence that it became overwhelming. The next 30 minutes does the opposite. It gives you the most optimistic view possible: apocalypse or optimism. Promise · 26 wordsAnd we here in the Senate and in the executive branch and in the House have a responsibility to reasonably, responsibly regulate in a timely way. Claims · 30 wordsThere are so many dangerous technologies developed in the last century where we put guardrails around them, whether it was jet airplanes or nuclear energy, automobiles or the banking system. Anecdote · 103 wordsI was in a terrible car accident end of June.⁵ I am here today because of airbags and seatbelts. Those were first advocated by safety folks like Ralph Nader. The Big Three auto companies mostly resisted. Why do we need seatbelts? When airbags came out, why the extra cost? I was in a five-car accident where all five vehicles were totaled and every passenger walked away.⁶ And as I walked away, as I walked out of the hospital, I said to a friend: Thank God. What a miracle. And he turned to me and said: And thanks for the seatbelts and the airbags.Anecdote Reasoning · 78 wordsWhen you get on an airplane and fly across the country, none of us probably can explain exactly how it works, exactly what the technology is. But we know that it has been maintained. It has been designed, maintained, funded, and flown by a tightly regulated group of people. Airline pilots operate under very strict standards. Those who maintain and operate airplanes operate under very strict standards. That is why we don’t have tragic disasters all the time. Claims · 181 wordsSo there are some good, concrete, bipartisan ideas for how we can inject some responsibility: some third-party auditing; some traceable records of what is happening, what has been done; some predeployment testing. Earlier this week, I introduced my bipartisan AI Systems Transparency Act with Senators Lankford and Britt and Schatz; ⁷four Senators--two Republicans, two Democrats, from very different backgrounds-- introducing a bill to require transparency about what is driving these AI models so we can better study them, test them, and understand them. There are lots of other good ideas that other Members have, but we, unfortunately, have a President who has said that the risks of AI are a hoax and, ⁸while hosting the titans of the AI industry, has refused to take seriously the idea that we in any way should meaningfully regulate this industry. I understand we are in a dire competition with China. I agree with the assertion that the society or country that wins the race to frontier AI will have a decided advantage in national security, in science, in industry, in business of all kinds. Reasoning · 26 wordsBut we can and should do both: compete against China, invest in frontier models, and have responsible safety to prevent the worst scenarios from coming true. Claims · 112 wordsWe are able to do this in banking, to give one other example. Delaware, where I am from, is a major banking center. Every one of the largest banks in America has major operations in Delaware, ⁹and they operate under some real restrictions. There are OCC and FDIC inspectors, regulators in every bank. They are required to follow know-your-customer laws, anti-money laundering standards.¹⁰ And after the 2008 to 2009 crisis, no one from the banking industry comes to me and says: We need a lighter touch regulation. They understand that safety and soundness--the rule of law--allows the United States to have deep capital markets and safety. Artificial intelligence is transforming our world, Promise · 73 wordsand whether it is adopting the AI whistleblower app that Senator Grassley and I are advocating for or it is adopting the AI Systems Transparency Act that Senator Britt and I and other colleagues are advocating for or other bipartisan bills, when we return in what will likely be a period of greater anxiety about AI, we need to act. We need to take up legislation and move forward reasonable and responsible regulation.No reason given