▶ 0:20:18I just the committee will come to order. Actually, I do have the subcommittee will come to order. Without objection, the chair is recognized to declare recess at any time. We welcome everyone here today for a hearing on the future of AI policy.
▶ 0:20:40I will note that this will be perhaps the last in a long series of AI hearings before several pieces of legislation will be marked up. So I encourage members on both sides of the aisle to make sure that that this panel of witnesses are asked questions that may be gerine to proposed legislation or legislation already offered.
▶ 0:21:05I now recognize myself for an opening Literally a generation ago or in technology 10 generations ago, a sharp young man graduated from Cal State San Marcus in my congressional He he joined a company that I then was CEO of Directed Electronics which had a inherent inventory
▶ 0:21:35problem. The inventory problem was that we had a few SKUs that sold well and we managed and hundreds of of SKUs that were constantly either out of or over supply of. It wasn't anyone's fault. We had simply grown quickly and there was a certain amount of inconsistency in what was being sold in a given month.
▶ 0:22:01So that bright young man took all of that inventory and the records and put it into SuperCalc, a precursor to Microsoft Excel. Within weeks, we had reduced our out of stock, increased our same day delivery, trimmed inventory to a level that actually saved us over a million dollars a year in inventory maintenance cost.
▶ 0:22:29That bright young man continued to work at the company for many years. He did not continue to use supercal for long because technology quickly gave better and better tools. The man, the program, the machine both are necessary to implement and make AI a But it was the man who made the machine that made the man a success.
▶ 0:22:56Over the last century, the US has led the way in virtually every area of technology because of our pro- innovation bias. We are the innovators. Well, China are the duplicators and Europe, yes, are the regulators.
▶ 0:23:17As we speak though, my home state of with an economy larger than that of Italy is rivaling the uh uh the European Union when it comes to trying to lead on This wouldn't be such an ironic uh occurrence except we are the home of innovation and yes, the new bastion of regulation.
▶ 0:23:42But just as in the 1990s when America led the AI the uh internet a light touch such as that offered by the president in his initiative in fact must be the direction we go.
▶ 0:24:02Anything else will give us a problem that I will If we in fact are not innovating 10 times faster than we're regulating, if the speed of innovation in the US is not at least months or if possible years ahead of China, their speed of duplication, some of it actually using AI to duplicate uh what we're doing will in fact cause us to lose our edge.
▶ 0:24:31My home state is part of the problem. The European Union is part of the problem. But as you'll see from our witnesses today, all 50 states have implemented some form of AI regulation.
▶ 0:24:44And in fact, there are in the neighborhood of a thousand pieces of legislation spread over 50 states that will create, if allowed to continue, a patchwork of indecision by the AI industry given conflicting regulations giving the inability to roll out with certainty uh technology, that technology simply will not be a priority. Let there be no do doubt though.
▶ 0:25:12Either we win in innovation and we win in AI or we lose our edge on the international stage. Vice President Vance said it best. America's a AI technology must remain the gold standard worldwide. We must continue to produce the next generation AI and we cannot do it with a patchwork of conflicting state laws. As of now, we are ahead.
▶ 0:25:43Let there be no doubt. We are ahead in hardware. We are ahead in software development. We are also on the edge of the leading edge of having the solutions for the energy problems. That includes modular nuclear reactors. It includes a willingness to provide uh innovative solutions. During the last break, I went to uh one of Apple's facilities, almost 1,700 acres located near Sparks, Nevada.
▶ 0:26:13What I saw there were some of the most simple buildings filled with endless rows of uh of various levels of chips for both AI and conventional data storage. What I also saw was a system that used zero conventional air conditioning to maintain that cooling.
▶ 0:26:38They had managed to beat one of the major causes of produ of of unrelated uh energy consumption which was air conditioning through an innovative system of evaporative coolers from locally available water and uh a filtration system that allowed those oper those to operate 24/7 without in any way being damaged by the high flow of air. they are making advances.
▶ 0:27:08This is over and above the innovation that we're seeing in chips and the ones that we plan to and the additional power. I'm going to contrast just over the border from California uh this location in Sparks, Nevada, because it has 64 gawatts of of generation power. Why would they need it? Well, they'd only need it in case of a power failure. Not so.
▶ 0:27:35The first time all were operational was to prevent blackout in California because California lacked the power. And by there going offline, Nevada was able to export power into my home state. That tells you a lot of the innovation in California, but not the ability to have those great new centers located there.
▶ 0:27:57In fact, Virginia just a mile from here is the number one location and other states are competing aggressively for it and if nothing changes they will win. These new laws will also affect early stage development because technical experts let alone start uh uh lawmakers are not capable today of predicting where we will be tomorrow. earlier this year.
▶ 0:28:27In fact, over the thinking on AI development and power needed took a sharp change and everyone on both sides of the Atlantic and Pacific are learning from what was released and that will continue. Of course, I don't want to be just a na naysayer because in fact, I am from an innovative state.
▶ 0:28:53I am from a state that is second to none in finding the best and the brightest in bringing them here. And although this committee does not have the d this subcommittee does not have jurisdiction over immigration, I want to make it clear here that AI development will also be about this committee working in a bipartisan basis to find ways to to not just attract but to retain the best and the brightest for that development.
▶ 0:29:19Let there be no doubt there are 300,000 Chinese students studying in America and most of them are being told to come home and bring with them what they're learning here. The release of the new AI action plan signal to the world that the Trump administration needs Congress to legislate America first AI. Now I know that sounds projorative but it isn't.
▶ 0:29:44The fact is that whether it was catching up on the uh if you will the interstate superighway under Al Gore senior or it was in it was leading on taking the arponet and turning it into the internet. We have worked together in the past. We have worked to limit states and to limit our restrain our own overregulation for the benefit of our economy and it has worked.
▶ 0:30:14I want to welcome the president's leadership and look forward to working to again to promote it. I want to additionally say that we have partners on both sides of the aisle. this committee uh including she's not here right now but uh uh Zolofrren and others have been great partners in this in the past and I expect they will be again I just want to leave us with one truism America has innovated and out innovated the
▶ 0:30:44countries around the world for Europe has become a regulator and an admirer of our our technology technology without embracing the way you get it. And China has become the most efficient stealer of technology and there the term duplication if it was truly innovation would be a compliment but it isn't. With that I recognize the ranking member for his opening statement.
▶ 0:31:14Thank you uh chairman Isa and thank you to the witnesses for your testimony when I drive from Georgia to Washington DC about once every six months or so and I get a chance to listen to the radio and uh scan and listen to all my favorite uh stations and and tunes. Uh you know, I've got to pass through three states.
▶ 0:31:40I go through Virginia, go through South Carolina, go through North Carolina, and then I hit uh Georgia. And um we all have have had experience in crossing state lines before and uh you'll have different speed limits and different levels of enforcement.
▶ 0:32:03you have um um the experience of figuring out uh which what are the rules or in other words cuz you're you're going 70 you're going 79 miles an hour at the state line and then flip to another state and boom all of a sudden the speed limit is 65.
▶ 0:32:25So, you've got to then do, you know, 74 to stop from getting pulled over and hope that uh that officers won't uh stop you. But when some suggested earlier uh that um we in Congress should preempt all state AI laws, they would not just have done away with states nent generative AI consumer protections.
▶ 0:32:55They would have preempted common law causes of action against AI companies as well. When the doctrine of caveat impour or buyer beware ruled American juristp prudence, consumers had minimal protection and were expected to thoroughly inspect products themselves. Judicial interpretations began to change in the early middle 20th century.
▶ 0:33:23As products became less straightforward and more complicated, common law developed to better protect consumers in products liability and negligence cases. Today, most Americans can hardly imagine taking apart a toaster, let alone an AI chat box to make sure that it works correctly.
▶ 0:33:45But caveat impour is effectively what advocates of a moratorum are suggesting we revert to when we talk about an AI When you preempt an entire field of law, you've you're preempting the common law right around right along with it.
▶ 0:34:06Supreme Court law has repeatedly found, as it did in Regal versus Metronic, that a federal laws reference to a state's requirements includes its common law duties.
▶ 0:34:24And in plain language, that means if Congress preempts state AI laws, we also preempt state common law unless the legislation explicitly says something else. Common law cases to protect consumers are already being filed against generative AI platforms.
▶ 0:34:46Two days ago, Senator Holly held a hearing on the harms to children using AI technology, calling witnesses whose children died or were hospitalized after interacting with artificial intelligence chat box. I know some of the parents and families are in the room today.
▶ 0:35:08Kristen Mc Kristen Bride, Juliana Arnold, Mandy Fernis, and Megan Garcia. I'm so sorry for what you all have been through, and I admire your commitment to Common law is crucial to the protection of Americans because it exists no matter whether there are comprehensive state laws on the books or no laws governing new technology on the books.
▶ 0:35:39Even when there are no statutes, common law helps us set a floor for a standard of care as a society. When some of my colleagues across the aisle talk about a moratorum preempting common law is exactly what they're talking about. Carveouts might be offered for some areas of the law.
▶ 0:36:03Others may get a loose regulatory structure, but what many don't realize is that the glue that holds the law together would be wiped out in almost every scenario. By protecting common law, we can protect that floor that ensures every person harmed can seek to have their case heard before a court of law. This basic standard of care can spur innovation by preventing a race to the bottom.
▶ 0:36:33And it can offer a level of security as the federal government and states determine what the best next steps are for AI in the United States. And with that, um, Mr. uh, chairman, I yield back.
▶ 0:36:47I thank the gentleman. Does a gentleman from California seek to be waved on to the committee? Uh, without objection, the gentleman be waved on, even though he's not a member of the self uh, subcommittee, and if others yield time to him, he'll be permitted to ask questions. Without objection, so ordered.
▶ 0:37:08It is now my pleasure notwithstanding the chairman and ranking members uh arriving uh to introduce our our distinguished panel of witnesses. Dr. David Bray is a distinguished fellow and chair of the accelerator at the Alfred Lee Lumis innovative council at the Simpson Center.
▶ 0:37:30He previously served as IT chief at uh bioteterrorism prep preparation and response program at the CDC and in the intelligence community. Dr. Bray is the recipient of a joint civilian civil commenation award and national intelligence exceptional achievement medal. Welcome. Mr. D Mr. Kevin Fraser is the AI innovation and law fellow at the University of Texas Law School.
▶ 0:37:59His research focus is on how to design regulatory regimes that increase adoption and use of AI. Mr. Frasier also leads the AI innovation and law program which prepares students for careers related to artificial intelligence. Thank you for being here. Mr. Adam Thy is the senior fellow for technology and innovation at the R Street Institute, a free market think tank.
▶ 0:38:30His work focuses on cultivating emerging technologies. He previously was senior fellow at the Makus Center and was president of the progress and freedom foundation. Professor Neil Richards is is the cop distinguished professor at Washington University School of Law where he also co-directs the Cordell Institute for Policy in Medicine and Law.
▶ 0:38:59His focus his work focuses on privacy law and information law and freedom of expression. We welcome all of our witnesses here. And as you may have seen on C-SPAN, it is the rule of the committee that all witnesses be sworn in. Would you please rise to take the and raise your right hand? Do you solemnly swear or affirm on a penalty of perjury that the testimony you're about to give will be true and correct to the best of your knowledge, information, and belief?
▶ 0:39:30So, help you God. Please be seated. Let the record uh reflect that all witnesses answered in the affirmative. As you also have heard many times, your true entire statement, including reasonable, even if expansive uh additional information you submit, will be placed in the record.
▶ 0:39:55So, as a result, if you're going to go past five minutes, do so by extension. and summarize what you do so we can leave time for questions. And with that, we will begin right they've got me. Mr. I want to make sure I get the right answer. Dr. Bray, after this it gets easy. We just go
▶ 0:40:18Thank you, Chairman Isa, Ranking Member Johnson, and members of the committee. I appreciate the opportunity to testify today. I'm Dr. David Bray, chair of the Lumis Accelerator at the Simpson Center, principal at Lead Do Adapt Ventures, senior adviser at General Catalyst Institute and a fellow with the National Academy of Public Administration. I work on tech, data, and geopolitical issues to help startups scale, communities adapt, and legacy organizations transform themselves amidst rapid global changes.
▶ 0:40:45My testimony focuses on advancing reliable, trustworthy AI consistent with the values of free societies and free markets. From these perspectives, I place my remarks in the context that the United States is experiencing multiple tech revolutions in addition to AI, advances in space technology, biotech, quantum tech, and the miniaturaturization of sensors and robots, all impacting US companies, our workforce, and our communities. With respect to AI, I would like to mention three noteworthy advances to inform our discussion.
▶ 0:41:14First, active inference AI models themselves demonstrate faster learning and use less data and less energy. Such approaches can be bounded by spatial or temporal limitations in ways that are human readable and interpretable across AI systems. Each of us as individuals could in the future restrict what AI systems do on our behalf.
▶ 0:41:34Second, openweight AI models with open source code have shown we can transform currently complicated processes such as a veterans affairs form into a conversational interface dramatically reducing the time to complete and speeding access to care. Third, federated learning allows AI systems to learn on data sets where they exist with proper consent, empowering both individuals and organizations to choose if their data sets and intellectual property are usable by an AI and negotiate a beneficial contractual relationship in return.
▶ 0:42:03Given these advances, three guiding principles drive my recommendations to the subcommittee. First principle, US strategies for advancing AI should recognize interdependencies between AI and other tech advancements. This requires a light touch policy framework. Recently, the National Academy of Public Administration has illuminated methods for sufficiently agile policy approaches to achieve measurable goals at the pace necessary given global changes. Second principle, different AI methods carry different risk and benefits.
▶ 0:42:32For example, AI approaches to computer vision and expert systems follow predictable outcomes whereas generative AI produces less predictable results. As such, AI policy should reflect these differences in AI methods. Third principle, there have been multiple waves of AI improvements over the years. We should expect continue advancements, which means US policy approaches must adapt accordingly. For example, the Stemson Center's Lumis Council intentionally brings together industry leaders to adapt projects to new AI developments.
▶ 0:43:03Even with different AI methods and the need for continuous adaptation, groups tied to specific domain applications of AI, for example, healthcare, transportation, and finance can promote data level interoperability across AI systems, avoiding silos. When electronic health record systems advanced in the 2000s, the United States encouraged the nonprofit health level 7 to evolve an open standard framework for interoperable clinical data with privacy controls. We should do something similar now for health and AI.
▶ 0:43:33We each deserve a choice as to when an AI uses our data and medical doctors should not be hindered by non-interopable AI systems. Given these principles, my recommendations are as follows. First recommendation, our principles and policies should help advance freedom, human agency, and individual liberties. We face global competition from the Chinese Communist Party regarding AI's future, including their AI plus initiative.
▶ 0:43:56US AI strategy must simultaneously encourage the advancements of the entire US AI industry and encourage the industry to advance individual freedoms. Second recommendation, upgrading existing domain specific laws is more pragmatic than adapting new sweeping AI regulations. I recommend a domain specific approach because the risk of different AI methods vary by applications.
▶ 0:44:18Examples include updating the Privacy Act of 1974, revisiting HIPPA, and reviewing other existing laws where the speed, scale, and scope of AI methods impact different risk calculus. Congress's recent efforts to upgrade banking laws with respect to stable coins is another example of updating existing statutes given new technologies. Third recommendation, assess what actions consistent with the US values of freedom, human agency, and individual liberties may need light touch policy to ensure AI advances freedoms across our nation.
▶ 0:44:48When updating policies that already exist, we should build on Justice Brandise's concept of a right to be left alone as law-abiding citizens, including choices about when personal data sets are and are not used by an AI, as well as when AI and any associated intellectual property shared is processed locally as opposed to a cloud-based instance. Any national AI strategy should ensure we do not stifle advancements towards reliable, trustworthy AI consistent with the values of both free societies and free markets. Thank you, and I look forward to your questions.
▶ 0:45:18Thank you, Mr. Fraser.
▶ 0:45:21Chairman Isa, Ranking Member Johnson, and distinguished members of the committee. Thank you for the opportunity to testify. My name is Kevin Frasier. I'm the AI innovation and law fellow at the University of Texas School of Law. Outside of teaching students, I believe there is no greater purpose for academics than sharing knowledge with policy makers. This purpose is all the more paramount when it comes to complex and novel challenges like harnessing AI to unleash human flourishing. A few months ago, Dr. Jensen came before this committee and announced that the nation that leads in AI will shape the future. Nothing has changed in the interim.
▶ 0:45:51What remains uncertain, however, is whether the US will retain its leading position. My goal today is to address the proper role of the states and the federal government in shaping AI policy on governing the use of AI. The 10th amendment reserves extensive authority to the states to regulate within their borders. On the matter of AI development, the founders offered their answer in abandoning the Articles of Confederation and adopting a strong centralized government capable of protecting and advancing the national interest.
▶ 0:46:17As I'll explain in the rest of my testimony, the founders infused three principles into our Constitution that when applied to the AI discussion, resolve debates about the authority of each actor to shape AI development. Subsequent changes in related areas of the law, namely the commerce clause, have given rise to the false impression that muddled judicial interpretation, somehow relaxed these principles. However, they remain as foundational today as they were 200 years ago.
▶ 0:46:43Adherence to these principles is essential both as a matter of fidelity to the founders's vision as well as to securing an AI regulatory posture that aligns with our federal system. The first principle is that the federal government alone is responsible for matters that implicate the economic and political stability of our country. The emerging threats to national security and economic stability posed by advances in AI place regulation of training frontier AI models squarely in the authority of the national government.
▶ 0:47:09To focus on one of many examples, AI has lowered the barriers to the creation and deployment of boweapons by bad actors. Defensive measures have not progressed at the same rate. Experts warned that with with significant technical progress, the nation would still need to adopt extreme measures to ready ourselves for a near future in which synthetic pathogens go undetected. That effort will flounder with second rate AI.
▶ 0:47:32Training frontier A model AI models and by extension safeguarding our national health and prosperity cannot be way by state laws, no matter how well-intentioned. Second, the extensive authorities reserved to each state and at their respective borders. As the Supreme Court has specified on multiple occasions, the equal sovereignty of the states is a fundamental principle of our constitution. Our constitutional order does not condone one state to intentionally and substantially interfere with the liberty and freedom of another.
▶ 0:47:58Political clout, economic might, nor population grants one state the authority to project its legislation into another. Whether a state is the fourth largest economy in the world or the 104th largest has no bearing on its authority to shape the lives of Americans beyond its borders. Though the Supreme Court has tolerated the inevitability of some regulatory spillover, its recent holding in National Pork Product uh Producers Council VROSS does not permit the sorts of regulations pending before many state legislators.
▶ 0:48:26Regulations that may deny all Americans access to a good itself because of the preferences of one political community. Building new pig pens to satisfy the preferences of Californians is technically and financially feasible. Training two AI frontier models, one to comply with the preferences of a single state and one for the rest of us, is a billion-dollar undertaking that rests on uncertain and evolving realities. Contradictory and vague state laws that impact AI development may thwart the sort of technological progress that has long fueled the American dream.
▶ 0:48:55under a patchwork of state laws that impact AI development. Uh we will see that Americans may never experience the improvements in education and healthcare that could have been realized by a national approach to pursuing the AI frontier. The third principle is that the ultimate authority in our constitutional system rests with the people. Our founders aspired for every American to exercise meaningful control over their daily lives. Extr territorial regulation of AI jeopardizes these and other features of individual agency.
▶ 0:49:24The nature of AI development means that if labs are compelled to comply with one state's regulations for model training, those requirements will be unposed on the rest of the country, rendering us all less likely to realize the benefits of AI advances. Americans may be able to move as freely as they'd like, but they would still find themselves using AI tailored by state legislators over which they have no control. Such a world is the antithesis of liberty. Denial or delay of the most sophisticated AI as a result of flawed state legislation is not a matter of mere inconvenience.
▶ 0:49:55It's a question of access to the greatest driver of human flourishing we've yet to develop. Thank you again for inviting me here today. I look forward to your
▶ 0:50:04Thank you Mr. The
▶ 0:50:06Chairman Isa Ranking Member Johnson and members of the subcommittee. Thank you for the invitation to participate in this hearing. My name is Adam Thear and I'm a senior fellow at the R Street Institute where I cover emerging technology policy. My message here today boils down to one simple point. Congress needs to act promptly to formulate a clear national policy framework for artificial intelligence to ensure our nation is prepared to win the computational revolution.
▶ 0:50:30If we get this wrong, the consequences could be profound in terms of geopolitical competitiveness, national security, economic growth, small business innovation, and human flourishing. Unfortunately, America's AI innovators are currently facing the prospect of many state governments supporting European style technocratic regulatory policies across America. As you noted, Mr. Chairman, more than 1,000 AI related bills are already pending across the nation. But some states are far more aggressive and influential on national market outcomes than others.
▶ 0:50:59Almost 50 AI related laws are currently pending in the in California, and New York is currently considering almost triple that number. Sacramento and Albany should not be dictating AI policy for the entire nation. That approach is especially problematic for so-called little tech innovators who will struggle with confusing, costly compliance burdens.
▶ 0:51:17America would not have become the global leader in digital technology it has if we had had 50 state computer bureaus or even a single California computer commission allowing to license every single aspect of interstate computing and treat the internet as a regulated utility. Thankfully, America avoided that fate because of wise bipartisan decisions that this Congress made in the 1990s, which let digital technology be be born free as opposed to being born into a regulatory cage.
▶ 0:51:46Laws like the Telecommunications Act of 1996 and the Internet Tax Freedom Act of 1998 included important provisions preempting and facilitating a national digital marketplace. The US is now the global leader in almost every segment of computing and digital commerce, thanks to this wise policy approach. Now is the time for Congress to work the same magic for AI by creating a national framework to prevent a patchwork of state mandates from undermining AI innovation.
▶ 0:52:13Colorado Governor Jared Polus has called upon Congress to preempt state AI laws such as the one his own state passed last year, and he has even endorsed the idea of a state AI regulatory moratorium like the one Congress considered this summer. Other governors have raised similar concerns. Connecticut Governor Ned Lamont has warned of quote, "Every state going out and doing their own thing, a patchwork quilt of regulations," he said. And just last week, New York Governor Kathy Hol noted how, quote, "It's hard when one state has a set of rules, another state does, and another state.
▶ 0:52:43I don't think that's a model for inspiring innovation." End quote. Congress could again try to implement a moratorum or it could formally preempt specific state and local regulatory enactments that impose an undue burden on interstate algorithmic commerce. If Congress chooses the latter option, federal lawmakers should first preempt state regulations of AI frontier models because the cost associated with such regulations would outweigh any local benefits. Such rules would create spillovers and undermine development of the system the nations need the nation needs to compete globally.
▶ 0:53:14State officials also lack the technical expertise and information about national security matters that could be relevant to AI safety considerations. Second, for issues related to so-called algorithmic bias or AI discrimination, Congress should preempt state efforts to regulate the development of AI systems and applications through cumbersome and confusing mechanisms such as AI audits or algorithmic impact assessments. To the extent any such ex anti-regulations are imposed, it should be done at the federal level and existing federal civil rights laws and non-discrimination standards should apply.
▶ 0:53:45Finally, Congress should also require the National Institute of Standards and Technology and the new Center for AI standards and innovation within NIST to oversee a new standing AI working group to coordinate and work to resolve other federal state AI policy matters. NIST and Cassie could develop uh help devise more workable consistent standards for AI policy matters. not already preempted by federal law.
▶ 0:54:07Even where the scoping of federal preeemption proves difficult, everyone should agree that AI development will be discouraged if America has dozens of different definitions of key concepts. Inconsistent standards will undermine market certainty and hurt investment, innovation, and competition. Ongoing congressional oversight of this process will be essential and Congress can simultaneously consider what sort of new light touch rules might be necessary at the federal level to address various AI safety concerns.
▶ 0:54:36Meanwhile, state governments still have a role to play and will have plenty of room to act using a diverse policy toolkit of generally applicable laws to address any real world harms that might come about from AI applications. In closing, the time has come for Congress to exercise its constitutional responsibility to protect the interstate marketplace and the national interest in the development of robust AI capabilities that will ensure the United States remains at the forefront of this technological revolution. Thank you for holding this hearing and I look forward to any questions you may have.
▶ 0:55:06Thank you and particularly thank you for mentioning our former colleague uh Mr. Polus, Professor Richards,
▶ 0:55:13Chairman Isa, Ranking Member Johnson, and distinguished members of the committee. Thank you for the offer.
▶ 0:55:18Uh, could you Yeah. Either put it closer or turn it on or both.
▶ 0:55:22Sorry, Mr. Sherman.
▶ 0:55:23Fantastic. Not a problem at all. It's We're talking tech here, so you know, let's let's go we'll go we'll go high-tech and turn them on.
▶ 0:55:29We we we have the automatic ones in St. Louis. So,
▶ 0:55:32of course you do.
▶ 0:55:35Chairman Iser, Ranking Member Johnson, and distinguished members of the committee. Thank you for the opportunity to appear before you this morning. My name is Neil Richards and I am the Koch Distinguished Professor in Law at Washington University in St. Louis where I direct the Cordell Institute. This hearing is about whether Congress should consider preempting state laws that touch on artificial intelligence technologies.
▶ 0:55:55And it is my firm and considered opinion that denying states the ability to regulate novel technology issues going forward would be a grievous and avoidable error that would not be in the best interests of American industry or the American people. I'd like to offer three highle points this morning. First, federal preeemption of state laws touching AI would be reckless and expose consumers to great risk of harm. Artificial intelligence, as we've already heard this morning, is not just one technology.
▶ 0:56:25It is a cluster of related and changing technologies that would be nearly impossible for a general preeemption law to define with care. In addition, AI technologies will likely affect every aspect of human life, just as industrialization did starting in the 19th century and as the internet did starting in the 20th.
▶ 0:56:45Like those before them, AI technologies will produce many good things, but also many bad ones, like kids becoming emotionally dependent upon chat bots, generative AI hallucinations affecting our courts by making up false citations, new ways to hack systems, and other harms critically that we cannot foresee today. At a time when we cannot be sure what harms will result, I'm sorry, at a time when we can be sure that harms will result, but we cannot be sure how.
▶ 0:57:14Depriving states of the ability to adapt to and try to mitigate these harms would be to disregard a clear and obvious risk and that is the legal definition of recklessness. Second, states have been pioneers of sensible tech regulation over the past three decades that has built essential digital trust for tech companies.
▶ 0:57:34If states had been banned from regulating the internet in 2000, we would have no broad requirement for website privacy policies, no databach notification laws, no laws banning employees from demanding the social media passwords of their employees, no laws regulating facial recognition technology, no substantive data security statutes, no comprehensive privacy statutes, no laws regulate preventing kids from accessing hardcore digital pornography or other dangerous content.
▶ 0:58:03No laws limiting the ability of tech companies to pedal addictive business models to children and much less enforcement of digital fraud, abuse, crime, hacking, and data breaches. Guided by these state legal guard rails in place to secure essential consumer trust, the past 30 years have seen the explosive success of Silicon Valley. Without state privacy and security laws, for example, we would still be afraid to give our credit card numbers to Amazon.
▶ 0:58:35State digital laws have tamed the worst excesses of the internet and helps to make it a trustworthy place for innovation, connection, free expression, and business. And broad AI preeemption would have the opposite effect for artificial intelligence. Third, I would like to address the claim frequently made by industry that state regulations somehow stifle innovation. As history makes clear, these arguments are, in my opinion, mistaken and misguided.
▶ 0:59:05Law creates and enables innovation by stabilizing the marketplace. It sets the ground rules for fair and robust competition, making the market safe and sustainable for consumers. Contrary to its libertarian origin myth, Silicon Valley was shaped by laws from the beginning.
▶ 0:59:22From government defense contracts to intellectual property laws and from securities laws to federal and state prohibitions on unfair and deceptive trade practices, law has always played a role in preventing scammers and thieves and in shaping corporate business practices so that they benefit society as a whole. It is the presence of regulation, including state regulation, that has led to America being a leader in digital technologies and services.
▶ 0:59:49While we can certainly, and I'm sure we will this morning, debate how much regulation and what kind is appropriate, having no new regulations at a time of rapid change would be a disaster. And if innovation is as magical as industry says it is, it can still do good things while respecting the policy choice of the people's elected representatives. In this way, the necessity required by reasonable regulation has been and should continue to be the mother of invention.
▶ 1:00:18In conclusion, stripping our states of any power to regulate AI, potentially anything done with a computer, would be a reckless and grievous error. State regulations have always played an essential role in building consumer trust and shaping the digital revolution for the better. Thank you, and I welcome your questions.
▶ 1:00:36Thank you. I understand the ranking member would like to make an opening
▶ 1:00:43if that's all right with you, Mr.
▶ 1:00:44It is always good to hear from you.
▶ 1:00:46Well, thank you kindly.
▶ 1:00:48Gentleman's recognized.
▶ 1:00:49I I'm very grateful to you for putting together this really important hearing and uh thanks to the witnesses for your statements. When commercially available AI debuted three years ago, the consequences were breathtaking from the start. Generative AI spurred scientific research and provided astonishing new tools to creators.
▶ 1:01:11The massive jumpstart to American innovation swept uh from social and economic domain to social and economic domain from pharmaceutical research and quantum computing to sound recording and film editing.
▶ 1:01:28But generative AI has also raised profound legal, practical, even philosophical problems such as whether individuals have a right to their name, their image, their likeness, and their voice when they're used in other people's deep fakes. If government um has unlimited power to engage in AI enabled surveillance of our citizens, and what are the appropriate standards of care, if any, for AI platforms to protect users against harmful consequences?
▶ 1:01:57While Congress takes the time to absorb the shock of these changes and these problems and examines the technology, the states have already begun to enact the first regulations on AI. We often talk about creation of rules for the road when crafting legislation to govern new technology. But I think this way of talking about consumer safety and technological ethics suggests that a road without speed limits would get us to where we're going faster. for generative AI.
▶ 1:02:25It's not about creating speed and safety laws and building highway guard rails, but rather building a road system in the first place. Some of my colleagues would argue that the construction of local roads is unnecessary. They say that without broad preeemption, without clearing the field of all state-based legal encumbrances, um AI companies and fledging startups will have trouble complying with state laws and will be put to a disadvantage. and then wither on the vine.
▶ 1:02:55I've heard little to suggest that broad preeemption is in fact the appropriate solution to this problem. Proponents of preeemption present Americans with a series of false choices, telling us we must choose choose a side between AI innovation or state powers and federalism, between business and consumers or national security and safe innovation.
▶ 1:03:19In 1816, Jefferson wrote, "Laws and institutions must go hand in hand with the progress of the human mind. As that becomes more developed, more enlightened as new discoveries are made, new truths discovered, and manners and opinions change with the change of circumstances, institutions must advance also to keep pace with the times." What we're seeing across America is individual states looking at these amazing technological developments and asking whether and how their laws need to change to protect their citizens and advance the
▶ 1:03:49common good. Um today you might think that the issue has some kind of necessary partisan veilance to it. Republicans on one side, Democrats on the other. But opposition to an AI moratorum I think is broad and bipartisan.
▶ 1:04:03In fact, when some of my Republican colleagues tried to pass a moratorium through our last spending bill, attorneys general from across the states, red, white, and blue, sent a letter to Congress saying, "Please don't do this to our state laws." In another letter, 17 Republican governors wrote to Majority Leader Thun and Speaker Johnson praising their quote big beautiful bill, but explaining that the moratorium provision stripping the right of any state to regulate this technology in any way without a thoughtful public
▶ 1:04:34debate was quote the antithesis of what our founders envisioned. I surely disagree with these governors on many things, but I think that they are right and should indeed be free to create what they call quote smart regulations of the AI industry that simultaneously protect consumers while also encouraging this everdeveloping and critical sector.
▶ 1:04:54In a statement submitted for this hearing, AI startup Bria wrote that a moratorum on state laws would create a giant vacuum and strip away the rules needed to quote raise capital, foreign partnerships, and build safely in order to win consumer trust. Without a road on which to travel forward, startups are cut out of the market in favor of large companies with the legal and fundraising teams necessary to deal with a barren legal landscape.
▶ 1:05:21Finally, some argue we need unrestrained AI development to properly compete with China. The subcommittees held many bipartisan hearings on the threat to innovation, AI supremacy, and IP from China. It would be amazing, I think, even dangerous to posit that we need to become more like China to compete with China.
▶ 1:05:41In fact, it seems more plausible to me that to believe that stronger, better products developed in America while protecting Americans and their data through American political processes and the passage of American laws will ensure that AI is both more advanced, more durable, and more internationally competitive. Protecting American innovation, investing in American research, developing American laws to deal with problems like deep fakes, political deep fakes.
▶ 1:06:10um discrimination through AI and so on and investing in our workforce I believe is the right way to win the so-called AI arms race. American safety is not at odds with AI innovation. That should be the baseline for any conversation we have about the best way moving forward. Um and I very much look forward to this conversation and I've already learned a lot from it. Yield back to you, Mr. Chairman.
▶ 1:06:33I thank the gentleman. All other uh opening statements be without objection all of their opening statements be placed in the record. Uh it is now my pleasure to go to the gentleman from Virginia, Mr. Klein, for five minutes.
▶ 1:06:47Thank you, Mr. Chairman. I want to thank the ranking member for referencing our third president, the gentleman from Virginia, Mr. Jefferson, and and one of our great inventors from our earliest days of our republic. and for the chairman referencing the work of our current governor, Governor Yncan, who is uh working to make sure that Virginia is the leader in in data center development. We continue to be the leader nationally and and we intend to stay that way as AI grows and develops. Mr.
▶ 1:07:13Frasier, you've written that the Constitution's intellectual property clause is first and foremost a directive to advance and spread knowledge. I think that was on X a couple weeks ago. Uh, how should Congress strike the right balance between protecting copyright owners and ensuring that AI regulation continues to promote the spread of knowledge consistent with that constitutional purpose?
▶ 1:07:31Thank you very much for the question, Congressman, and thanks for the follow or at least uh perhaps the like. Uh, I think in this regard, the constitution is clear that the IP clause has always been grounded in progressing the progress or excuse me, progressing the promotion of science and useful arts. And here if you go back to the founding articles as well as uh subsequent interpretation by the Supreme Court, the focus has always been on making sure that there's the spread of knowledge across the country.
▶ 1:07:58IP laws uh grant an exclusive right to creators to attempt to incentivize that creation. What we need is to make sure that there is an economic analysis of the extent to which those laws are working as intended. The purpose of the IP clause is not the profit of creators. It is the progress of society. And so what we need to get back to are those first principles when it comes to examining copyright law and patent law.
▶ 1:08:23Right now if you look at analysis from scholars such as Richard Watt, you will see that the preponderance of copyright benefits are not going to your average Joe and Jane author but to large publishers and so we need more analysis on that front. Would you agree that ensuring transparency in AI systems such as being able to trace what data is used to train a model is essential both for protecting IP rights and for maintaining public trust in AI
▶ 1:08:48I would agree that broad overviews of the sources of training data are important to get an understanding of where and how models are being trained. And in that same spirit, could giving creators a private right of action for tampering with content credentials help strike the right balance between protection and innovation?
▶ 1:09:06My own estimation is that granting that sort of right would be a significant uh barrier to AI innovation giving the given the centrality of access to data for innovation. We've seen that many courts and many scholars have regarded the use of data as a transformative po purpose under copyright law and denying the ability to train on wide swaths of data uh would be a real hindrance to our ability to leverage AI.
▶ 1:09:32As many folks have said on many occasions, uh bad data leads to bad AI. If you want better AI, you want better quality information. And if we throw many legal gears into that equation, we won't get the AI we deserve. Thank you, Dr. Bray. As we consider whether and how best to regulate AI platforms, do you believe that we must avoid the same mistakes we made in the early days of the internet with broad safe harbors that gave platforms a free pass for enabling copyright infringement and countless other harms?
▶ 1:10:02Thank you for that question, Representative Klein, and as a fellow Virginia, glad to be here. Um so I was around actually working in the 90s on the early days of the worldwide web and my my observations would be we did fit for purpose for the 90s.
▶ 1:10:19Now in the two decades since we have seen the rise of applications on top of that technology where we may need to make adjustments and I think what we need to separate is the desire to roll out the technology so that the entire nation could have access to the internet and at the same time if we see that the applications need adjustments for the law that would make more sense to to to adjust.
▶ 1:10:42However, I would say what we need to recognize as well as we go forward here, where are we trying to advance the technology so it can be used by startups? It can be used by communities. It can be used by legacy organizations that haven't gone AI native yet. And at the same time, if we see there are applications where we want to prevent harm to individuals, to children or things like that, adjust the applications while not limiting innovation on the technology.
▶ 1:11:06Just like other businesses, bad actors have to be accountable for the harms they cause. And in addition, if we consider some type of temporary pause for state level AI specific regulation, we have to ensure that other generally applicable state and federal laws continue to apply with copyright law being one example. Uh do you think AI platforms should be held to the same standard of accountability as any other business including when it comes to respecting copyright?
▶ 1:11:29Absolutely. And I would actually say that's why I'm so excited about federated learning because there actually could be the opportunity where whether you're a recording artist, you're a musician, you're an individual, you could actually say here's the data that I have pulled, you can learn on my data in situ as opposed to shipping it somewhere and we can have that actually recorded as a transaction and then in return I'm getting benefit whether it be financial or otherwise. It's a new model that's actually quite possible.
▶ 1:11:56It's been possible for more than five or six years and we can motivate people to do it. Great. Thank you. I yield back.
▶ 1:12:02Gentleman yields back. Who seeks recognition? The gentleman from Georgia is recognized for five minutes.
▶ 1:12:07Thank you, Mr. Chairman. Professor Richards, uh, I mentioned in my opening statement my concern that broad federal preeemption of state AI laws would also preempt common law causes of action. How does common law, particularly tort law, help protect Americans from harm? Well, thank you for the question, uh, Representative Johnson. Common law is the foundation of American law. It is older than the United States.
▶ 1:12:36It goes back to the to the colonies, to the English tradition. Common law has a flexibility to the law. If if we think about uh my own specialty, privacy law. Um, there was a reference earlier by one of the other witnesses to Justice Brandise's right to be let alone. Um, privacy law in America was originally a product of of common law. Uh, where the law adjusted to to realize that data about people, information was being collected without their without their willingness was being disclosed.
▶ 1:13:06And an important line of cases relevant to this subject today uh to protect the names or likenesses of people who pictures and names were being used to sell products without their consent. So common law is a is a tremendous source of flexibility and vitality in our law that allows the law to adjust to change circumstances like the advent of technological revolutions such as artificial
▶ 1:13:31Thank you. Can common law be used to protect Americans even in the absence of explicit statutes?
▶ 1:13:40Absolutely, sir. I mean that that is sort of the the nature of the common law that that judges can apply the existing rooted principles of of of tort law and contract law and property law and they can from those general principles divine specific applications that can can provide new protections so that the law continues as it always has to evolve alongside technological invention as uh the ranking member referred to uh Mr.
▶ 1:14:06Jefferson's uh statement from 1816, which as a as a proud graduate of UVA, I also endorse.
▶ 1:14:13Some of the current lawsuits against AI companies are being brought under common law to hold companies accountable for the harm that their products have caused to children. For example, Megan Garcia is suing Character Technologies and Google after her 14-year-old son, Su Set died by suicide.
▶ 1:14:36She testified before our colleagues in the Senate this week that his death was quote the result of prolonged abuse by AI chatbots, excuse me, was the result of prolonged abuse by AI chatbots on a platform called Character AI end quote. The chatbot sent Suexual messages and asked him to come home to me as soon as possible.
▶ 1:15:06Others have filed lawsuits against Character Technologies and Open AI for wrongful death, negligence, and other causes under both common law and state laws about deceptive or unfair trade practices. These tragic cases show some of the worst possible harms that can arise from AI technologies. Professor Richards, does an AI moratorum run the risk of impeding these lawsuits that seek to hold companies accountable?
▶ 1:15:37It would. I I I particularly if it were defined broadly and and let me also say in in response to your uh question uh Representative Johnson, um as a parent myself, my heart goes out to the to the families who have who've lost their children. Um I I think when we think about laws like negligence or rules like negligence, negligence was the great innovation of the common law to respond to industrialization.
▶ 1:16:03It means that anybody acting uh against other people must behave in a reasonable way and not cause unreasonable harm. I am sure that the car companies and the railway companies and the industrial companies of the 19th and 20th centuries would have argued that the common law developing the law of negligence would have impeded innovation.
▶ 1:16:22But actually, it safeguarded the development of those technologies by enabling us to be able to drive cars and ride on the rails and fly in airplanes and and otherwise enjoy the benefits of our inventions knowing that we are safe and we are protected. And where those technologies or their their deployers overstep the line, we have a right of action to defend our rights and protect our families.
▶ 1:16:43Thank you. And it was states that developed the common law. Professor Richards, beyond common law, are there some areas where it's appropriate for uh states to lead the way on laws about AI technologies? And if so, what sectors or use cases should continue to be the providence of the states? I I think the answer to that question is yes. And I think particularly where there is deployment of AI rather than generation of AI.
▶ 1:17:12um the the use of AI in in point of sale devices, um employment discrimination, consumer protection, the traditional provinces of of state regulation and and let me also if I could add one additional thing uh
▶ 1:17:30states uh have filled the gap that is that this Congress which did not regulate the internet, did not regulate privacy generally have done so. And I think with AI technologies, uh, if Congress is is for whatever reason not able to pass comprehensive legislation protecting Americans, states will continue to fill that gap as they have in the internet age.
▶ 1:17:53Thank you. Uh, we now go to the gentleman from Texas, Mr. Gooden, for 5
▶ 1:17:59Thank you, Mr. Chairman. Uh, Mr. as a trend, it seems like every state's jumping to regulate AI and perhaps some of them are doing it just to show early Do you think that's well thought out? And also, what are the long-term effects of having a decentralized patchwork of laws? How does this help or hinder new
▶ 1:18:20Thank you very much for the question, Representative. Always good to talk to a fellow Texan. Hookum. Uh so I think in uh any prior study we've seen of the impact of a rush to regulate among states there is a real noticeable impact on small businesses. If we look to research for example from engine uh done in conjunction with the Michigan Ford public policy school we see that just changing a privacy policy statement may be $6,000 in funds to outside council.
▶ 1:18:49That's $6,000 out of $55,000 of monthly and uh monthly revenue and operational expenses. So from a small business perspective, the rush to regulate is a real hindrance on innovation. I also think that the rush to regulate uh among the states creates a patchwork and a huge risk of extr territori extr territoriality in terms of application. Uh we've talked a lot about Virginia today which is welcome.
▶ 1:19:16At the time of the founding, Virginia had around 700,000 residents. Delaware and Rhode Island something around 30,000. The founders didn't say there was a Virginia privilege or we should have a Virginia effect. They did not want to see that happen. Instead, they made sure that states stayed in their respective borders when it came to regulation.
▶ 1:19:34Thank you. And is it possible, do you think, for bad actors to misuse inconsistencies, especially in terms of violating intellectual property rights? I think that we've seen a documentative effect of what's referred to as regulatory overload uh as folks at the Mercada Center have written about when we have endless litigation, endless labels, endless warnings. Actually, what we get is less safety because people don't know what law to adhere to. If you talk to a lot of startups today, they don't have a public policy person.
▶ 1:20:04They don't have a general counsel. So, just adding more laws to the equation actually reduces the odds of user
▶ 1:20:10Thank you. I appreciate that. I yield the balance. uh of my time to Mr. Korea from California.
▶ 1:20:16Thank you, Mr. Gooden. Gentlemen, listening to your debate today reminds me of what General Patton used to say, which is uh lead, follow, or get out of the way. AI is moving faster than we imagined or we even expected just last year, touching every aspect of our lives.
▶ 1:20:35Most of our constituents, like many of us here, don't know a lot about it, but they know enough to expect that we here will protect them, their jobs, their their intellectual property. And so the debate about whether it's local control or federal control, I think, is second to the fact that we just can't move on this stuff at the federal level. And so, Mr.
▶ 1:21:04Frasier, you're from Texas. I'm from California. Fourth largest economy in the world. But what do we how do you coordinate federal and state action to make sure that we respond to our constituents responsibly? Thank you.
▶ 1:21:22Thank you for the question. And I I was a Berkeley law grad uh so I share some California ties. I want to emphasize that we do need to see regulation in this space and we need to see that Americans are protected and especially our vulnerable communities and children. Uh what I'm concerned about are laws like AB 1046 out of California for example that imposes on AI companions a desire to uh prioritize factual accuracy over a user's preferences.
▶ 1:21:52Specific 1047 or 104
▶ 1:21:54uh AB 1046.
▶ 1:21:56Okay. prioritizing factual accuracies over uh the preferences of the user, in this case a child. To which I ask, who gets to define factual accuracy? Is it California state government? Who's going to answer the question of is Santa real for a seven-year-old? Who's going to make factual determinations about religion for that child user? Those questions shouldn't be answered by California for the rest of the country.
▶ 1:22:19I would ask you to also look at SB53 that's now being addressed in the California State Legislature and see what you'll find on that as well.
▶ 1:22:27I think SB53 is the least bad option I've seen with respect to AI development regulation as we've discussed in this hearing. AI development in my opinion and in the opinion of many should be left to the national government as a
▶ 1:22:41Anything you agree with on SP?
▶ 1:22:43I I very much agree with the whistleblower protections. I think that's an important mark. I also agree with the fact that it calls on regulators to revisit definitions and terms frequently to make sure they're working as intended.
▶ 1:22:54Thank you, Mr. Gooden. And Mr. Chairman, I yield.
▶ 1:22:58Thank you. We now go to the ranking member of the full committee, the gentleman from Maryland.
▶ 1:23:03Thank you very much, Mr. Chairman. Um, I don't understand the attack on the patchwork of laws. Um, I mean, first maybe that's because I'm a quilt person and I like patchworks, but um, but isn't that what federalism is? Federalism is a patchwork. Um, and that's how that's the glory of the American governmental system.
▶ 1:23:28um to be sure the the Congress gets it together to adopt a national law eventually on everything from the Clean Air Act to the Clean Water Act to the National Labor Relations Act, the Fair Labor Standards Act. But I think it would have made no sense to say um before those federal laws were passed, let's wipe out the state laws that exist on child labor or minimum wage or not polluting the water or not polluting the air.
▶ 1:23:56Um in fact that contradicts what I thought the central dynamic of federalism was which is the states are the laboratories of democracy as I I think that was Brandeise. Um and then the different changes that they make are compared to each other and then they bubble up and Congress takes all of it into account when uh it decides to attempt a nationwide approach. Is that a fair statement of uh the situation? Professor Richards.
▶ 1:24:25Uh, absolutely. I think the, uh, I I also have a degree in early American history from the University of Virginia. Um, my studies in early federal history.
▶ 1:24:35Virginia is getting a lot of play today. I don't know. As a Marylander, I got some questions about that. Yeah.
▶ 1:24:42Uh, AB to Would you like me to continue,
▶ 1:24:45please? Absolutely. the the goal of uh federalism is to have laws that are more responsive to the people who are closer to those representatives that that the legislatures of particular states can adapt to that particular state's problems and strengths and also to to protect to experiment.
▶ 1:25:03Okay. So in just to restate the obvious, there was an attempt to impose a moratorum on state laws or to wipe out state laws. How is that different from the way that federal preeemption has taken place in the other cases that came to mind for me?
▶ 1:25:22Well, sometimes federal preeemption can just through the supremacy clause preempt particular laws or or laws that are inconsistent with the federal mandate. In addition, Congress is allowed to operate in ways to set a general national standard, but still allow states to experiment with stronger standards so that the innovation in regulation could continue a pace with the innovation in technology as as you and Mr. Jefferson put it so well.
▶ 1:25:43Well, and and in addition to those differences, isn't it the case that a moratorum today would just wipe out state laws without substituting anything, without imposing a national law? I mean, absolutely. There's is there any precedent for just doing that, saying we don't want any state laws at all while we think it over or while we're stuck in some kind of legislative
▶ 1:26:05I can't think of one. That's why I think it would be disastrous because depending first on how the law is defined, it could sweep very very broadly and and take out laws that are that are important and protective that everybody and this panel would agree are good laws.
▶ 1:26:19And second, if you have uh broad preeemption, um this would be a defense that tech companies could make in every piece of litigation, increasing the cost in a litigation system as the contours of that preeemption definition could continue to affect litigation years into the future.
▶ 1:26:37Yeah, some people are with us today who um are involved in or have been involved in different kinds of litigation and uh my heart goes out to them being uh um a father who's lost a son. Um and these are all people who've lost children um in different kinds of interactions with um uh chat bots and other kinds of AI technology.
▶ 1:27:06And I just want to recognize Kristen Bride from Oregon who lost her son Carson. He was 16 in 2020. The same year we lost uh our son Tommy. Juliana Arnold who lost her 17-year-old daughter Coco uh to fentanyl poisoning after she purchased a counterfeit uh pill online.
▶ 1:27:27Megan Garcia from Florida um who lost uh her 14-year-old son Su who took his life in February after months of abusive interactions with a character AI chatbot. Um and Jane Doe from Texas um whose son JF suffered severe physical and mental health harm after multiple chatbots instructed and encouraged him to engage in self harm and self violence.
▶ 1:27:54Um, and so all of which is to say in my mind there are profound problems here that we really do need to deal with. And the last thing I would want to do is to try to nullify states that have already addressed the problem in response to constituents dealing with a nightmare like that without replacing it with something. And I'm not averse to the idea that there might be a national law that works, but certainly imposing a legislative vacuum on the country would be a really dangerous way to go. Thank you, Mr. Chairman. and I yield back to
▶ 1:28:24You're most welcome. And with that, we go to the gentle lady from Florida for 5
▶ 1:28:32Thank you, Mr. Chairman. Mr. Frasier, I'd like to return to you. I appreciate so much your discussion of the commerce clause, and I think you made some important distinctions for us already in your testimony when you talked about the need for the federal government to intervene and think about preeemption. when we're talking about a subject that affects the economic or political stability of the United States and you drew a distinction between pigsty and artificial intelligence.
▶ 1:29:02Would you elaborate for us please about why you believe the things that we're discussing here today do go to the heart of the economic and political stability of the United States and should be distinct from those areas where the laboratories of democracy concept actually works. Thank you for the question, Representative Lee. Uh, I think it's profoundly important to get back to that Brandeise quote about laboratories of democracy.
▶ 1:29:29There's a forgotten portion of that quote, which is without risk to the nation. You can run an experiment without risk to the nation. Many of these experiments that we're seeing proposed and enacted in California do pose a risk to the nation because they try to impede AI innovation itself. When we see individual states reach into the AI development process, they're not just tinkering with a modular process. There is not a AI specific training run for California that OpenAI does.
▶ 1:29:58Anthropic doesn't train its models 50 times over for each state. So while there may be a lane for state regulation, and I believe there's a lane for state regulation with respect to AI use, we have to follow up and ask the question of what does a real experiment look like? And that experiment can't be one that exceeds the borders of that state. And yet California's bills time and time again would uh result in labs having to change their practices the nation over. And I've lived in California. I've lived in Florida. I've lived in Texas.
▶ 1:30:28I've lived in Oregon. I've lived in Massachusetts and in DC. And I can tell you in each of those places, they don't want Californians dictating the terms of their AI.
▶ 1:30:36And I also need to follow up on this question. You said something interesting when you were talking about current copyright law. how it operates uh in really ensuring that we are still honoring the concepts of content creators and intellectual property. You suggested an economic analysis of what's happening with the use of this content and training models. Would you elaborate for us a bit more on how that would look and how we can get to the bottom of how to properly compensate uh those content
▶ 1:31:06Happily, thank you for the question. If you look at current settlements for example uh in the BART's vanthropic decision and you begin to analyze who are who may that those funds actually go toward uh a large number of that fee is going to go to publishers. It's not going to go to the actual authors themselves. So if we are trying to incentivize the creation of new art and new uh new discoveries and new scientific discoveries, copyright may not be the vehicle we need.
▶ 1:31:34It's not serving the same purpose it did in 1789, back when it was just limited to 14 years with the possibility of a 14-year renewal. It is now 70 years plus the life of the author. That is an incredibly long time, especially when you consider that the founders really hated monopolies. So, the fact that we've ended up in a world in which a handful of publishers may be able to dictate the quality of our AI, I think is anthetical to the original purpose of the IP clause.
▶ 1:32:01Thank you. And Mr. Sarah, uh, one of the things that you touched on was the idea that Congress could explore giving NIST or Cassie more authority to develop standards for AI frontier models. Would you share more on your perspective of how we might do that? Should we designate a single federal entity to try to develop those standards? Uh, share with us a little more on your thoughts there, please.
▶ 1:32:25Yeah, absolutely. Thank you for the question, Congresswoman. Let's be clear that the reason that NIS needs to play a role here in this new Cassie body is because they have the ability to address exact exactly what the problem is here, which is that many states are attempting to impose a very technocratic type of design on artificial intelligence models and systems preemptively in almost European style way. And that's a huge problem.
▶ 1:32:47I'll just again quote from Governor Jared Polus who said, "Government regulation that is applied at the state level in a patchwork across the country can have the effect to tamper innovation and deter competition in an open market. But it's not just that. These states lack the technical capability to do some of this in certain circumstances and they lack the information needed to do it properly. We have set up this body. I would sure I should remind the committee set up under President Biden and retained by President Trump in a bipartisan move and just renamed to focus more on standards and innovation. That's a good plan.
▶ 1:33:17Once again, we have a bipartisan agreement here. We've got a new technical body and it can handle it in conjunction with other existing policies, both federal and state.
▶ 1:33:25Thank you, Mr. Mr. Chairman, I yield
▶ 1:33:27Thank you. We now go to the gentle lady from California, Miss Lofrren.
▶ 1:33:32Uh thank you, Mr. Chairman. And um uh before I make remarks or questions, I'd like to ask unanimous consent uh to put into the record a letter from the California Privacy Protection Agency.
▶ 1:33:46Without objection, so ordered.
▶ 1:33:47Thank you. you know, um the title of this hearing, uh I take a little bit of exception to. Um you know, I get it. People bring up California because we set the pace. Uh but it's worth noting that I think my colleague from California, Mr. Koreah, mentioned California is the fourth largest economy in the world with over 4 past Japan last year.
▶ 1:34:17and trailed only Germany, China, and the United States as a whole. It's number one. California is the number one state for manufacturing. It's uh home to the most Fortune 500 uh companies of any state. Uh more than 40 are in Silicon Valley, my home. It also has the highest agricultural output of any state.
▶ 1:34:37It's home to the nation's um five of the nation's top 10 public universities, UCLA, uh UC Berkeley, UC San Diego, UC Davis, and UC Irvine. It accounts for over 12% of all university R&D expenditures in the United States with the University of California system alone uh spending more than 12.1 billion.
▶ 1:35:02and it receives more NIH funding than any other university system in the United States. Now, these aren't just vanity stats. They're the foundation of the modern innovation economy that California has built. World worldass universities, labs, investors, workers who turn ideas into jobs into growth.
▶ 1:35:27California has been and remains the leader in technology and the engine uh that built our economy uh now powers our AI leadership. California leads the world with 32 of the top 50 AI companies uh based here. So I do think although it's always fun to criticize the most successful state, we must be doing something right uh to have achieved all of this.
▶ 1:35:54Now this is a hearing on uh AI and I you know some of the comments made by the uh the witnesses I agree with. Mr. Bray uh you mentioned that upgrading existing domain specific laws is more pragmatic than attempting sweeping new regulations. I very much agree with that. But I think that's also going to need room for regulations that are specific or laws that are specific to each state.
▶ 1:36:25I mean, there are things that are the proper purview of states and there are things that are the proper purview of the federal government and certainly I was a critic of Mr. Weiner's bill from the last session in that it overreached in the uh national security uh era. Um I also agree Mr. Theer that the EU approach is incorrect. I mean to try and micromanage the workings of the AI system is doomed to failure it seems uh to me.
▶ 1:36:54Um however the recommendations that we simply preempt when we have nothing put together now is problematic and I've just got to say Mr. Chairman and you are also a member of the science committee along with me. We had a pretty effective bipartisan task force on AI in the last Congress chaired by uh Mr. Obernolulti from California as well as Ted Lou from California and they took the first step.
▶ 1:37:24I mean they didn't finish the job but they haven't even been reconstituted in this Congress. Um, we do heavily rely on NIST, an agency that is widely respected uh in the Congress and in the technology world, but we've got to look at what's happened to NIST.
▶ 1:37:43They've been eviscerated uh by the uh the Doge people and I failed to see how they're going to be able to perform uh the tasks that we're hoping that they can perform uh given what has happened to them. So, I would just uh like to say that uh we ought to be working uh on a bipartisan basis. Again, the science committee staff um staffed the task force.
▶ 1:38:10I would urge the speaker and whatever influence the chairman can have to reestablish that AI task force as a super subcommittee of the science committee um so that we can get more work done and get to where we need to be to have the guard rails and the standards that are appropriate at a national level while also recognizing there are things that are of value uh in in the state level.
▶ 1:38:36was the note from an anthropo uh a open IIA just mentioned Buffy Wix online age verification something they support you know so there are things that the states can do there's things that the federal government can do but we're not going to do anything unless we can get our act together and reinstitute that task force and get some more work done so with that Mr. Chairman, I yield back.
▶ 1:38:59I thank the gentle lady and and I I might note that both sides of the uh the title reflect California as the home of of the innovation that's driving it. Uh I might also take an opportunity to completely agree with your comments related to the need for us to act that a uh a uh preempting without a solution without some of the work that is being currently worked on both here in the House and Senate would not be well received.
▶ 1:39:29Uh so I I think we have the groundwork uh for a lot of uh the kinds of of work that you and I have done together. Uh, and I look forward to very much to this hearing being the beginning of us launching bipartisan legislation because we do need to act in some cases and you have always been a good a good partner in that acting. And with that, we continue on no no with deference to to my great state, the Commonwealth of Virginia.
▶ 1:39:56We'll continue with the Cal with the California effort and go to Mr. Kylie for five minutes.
▶ 1:40:03Thank you, Mr. Chair. And uh I wholeheartedly agree with my uh colleague uh from California that our state continues to be the center of breathtaking uh innovation uh worldwide. Uh however, the competency of our state government is another matter uh entirely.
▶ 1:40:19uh not to uh impugn the competence of any of my uh former colleagues in the Sacramento legislature, but this is a body that struggles with things like building roads, uh delivering uh electricity, managing forests, building dams, and getting water to come out of hoses.
▶ 1:40:37Uh so the notion that this is the right body uh to regulate the most powerful technology in human history uh whose workings are actually largely beyond the understanding even of the technologies creators uh is a fairly fantastical notion.
▶ 1:40:54Um you know not only that but uh we are also faced with a technology that continues to to uh to accelerate in capability uh in an exponential way in a way that is unlike anything we've seen before. Just to take one very specific example you have leading models have recently gotten uh the gold medal in the international math olympiad uh something that most experts thought was still going to be years away.
▶ 1:41:19And so I do think the risk that California is going to drive AI policy uh for the entire country is is a very real one. Uh and I think that a national framework uh that seeks to stop that from happening uh is is needed and appropriate. And more specifically, I see the federal role as uh including the following.
▶ 1:41:42I mean number one of course we need to be prepared to combat concrete harms as they arise and uh harms where uh the use of AI tools can sort of accentuate uh the risk. Uh secondly uh there needs to be uh you know uh risk assessment type tools. Uh maybe and as much as I'm giving California a hard time there are some decent ideas in this latest bill uh incident reporting uh uh transparency as far as safety protocols.
▶ 1:42:10Um, of course, there's a uh tremendous role uh for uh the federal government when it comes to the infrastructure needs uh behind the ever escalating uh investment uh in data centers. Um beyond that, I think it's very important that policymakers continue to be apprised as to uh the capabilities of these models. Uh in fact, both sides of it, the the risks as well as the capabilities. There are of course channels that exist uh both between the labs and government and among the labs themselves.
▶ 1:42:41Uh but I think that most of us as policy makers unless you're out looking for it are not kept up to speed on exactly uh where the leading edge is. I think that's all very important. I think there could also be a lot more uh investment in actually safety and alignment rate related research. Uh the labs do this themselves but they're not necessarily incentivized to do it. And so there could be more of a federal role for promoting basic cutting edge uh safety related and alignment related research.
▶ 1:43:10Uh but then finally and and maybe most importantly I think that uh part of this conversation that you know we've been focusing a lot on uh when it comes to discussions of AI there have been more of them happening here lately but uh they've been really oriented on the aspects of the issue that are familiar. Okay. So the issues related to energy, issues related to water, uh some of the risks that are of a familiar kind.
▶ 1:43:33Uh but the the discussion has not focused much on the broader question of how are we going to prepare society for the enormous changes that are likely uh to be ushered in uh in the coming years. And so when we get to this idea of states as laboratories of democracy or of experimentation, I think this actually is maybe uh the context in which that idea is most relevant.
▶ 1:43:56uh because when it comes to sort of regulating the capabilities and constraining the capabilities of the systems themselves, uh the laboratories of democracy idea isn't really fitting. Number one, because there is an enumerated federal power when it comes to interstate commerce. Number two, you talk about experimentation. This is sort of something that we have to get right and we only get one shot at.
▶ 1:44:17uh there's a widely shared view uh that once AI capability crosses a certain threshold whether that be recursive self-improvement or some other threshold there's sort of going to be an escape velocity uh so that has implications number one for the sort of uh narrower geopolitical context of which country leads in the technology but also for the broader idea of is this technology going to be aligned with and beneficial to uh humanity but I do think that states can play a role when it comes to uh preparing ing society
▶ 1:44:47for using this technology for good uh in uh various domains. So for example, education. I think that you're seeing states already uh experimenting with ways that uh AI can be used to close achievement gaps and to bring uh tools to students unlike anything we've ever had before. transportation states can take a lead and some states have taken a lead uh in uh preparing uh our transportation systems for the increasing uh capacity for autonomy uh within various uh modalities.
▶ 1:45:18Uh and then finally uh and and not finally there are various others examples but another final example I'll mention uh is the use of AI itself in government in order to improve government processes which we're seeing some of uh here at the federal level. We're seeing some experimentation with states uh in other countries across the world. But when it comes to being laboratories of democracy and the role of states here, I think that's probably where states can be most valuable and our role in Congress uh should be to pursue some sort of federal framework. I yield back.
▶ 1:45:46Thanks a gentleman. We now go to the patient gentle lady from North Carolina for her five minutes. Miss Ross.
▶ 1:45:52Uh thank you very much. U Mr. Chairman, I have a unanimous consent request. I ask unanimous consent to enter into the record a letter by Frank Cullen, executive director of the Council of Innovation Promotion to you and the ranking member dated September 18th, 2025, which expresses the council's concern regarding recent proposals for con uh for Congress to impose a moratorum on state level regulation of
▶ 1:46:19Without objection, so ordered. Um, thank you again um to both uh the chair and the ranking member for organizing this very important hearing and to the witnesses for your testimony. I'm glad that we're talking about how Congress and other lawmakers can responsibly legislate and regulate around AI. I represent the research triangle in North Carolina.
▶ 1:46:43I have seen the incredible things that AI can do particularly in the medical area and in um biioarma and I'm just just blown away by the powerful and positive use of AI.
▶ 1:46:59But I've also seen the negative effects of AI and I'm thrilled that the two one of those issues has brought been brought up by um Representative Klene and Representative Lee and that's the ongoing necessary litigation that's happening with content creators and copyright. I was with the head of anthropic this morning talking about how much money they're having to pay for what they did that was illegal. flat out illegal.
▶ 1:47:29And I hope and look forward to working with Mr. Chairman and the ranking member to make a hearing on that issue. And we've had a couple of those hearings happen again, but do it in a way that we can promote good behavior by AI companies.
▶ 1:47:48I also um I know I love California and I know we're talking about California, but I want to bring up some crucial areas where other states have regulated AI in necessary ways. And I know that we have the parents of children who have been hurt by AI here and the states are ahead of Congress in protecting our children.
▶ 1:48:13And given our inaction, many states have stepped up, passing legislation covering topics that run the gamut from expanding CESAM laws to cover AI generated material in Alabama to prohibiting AI from use being used to provide mental health care services in Nevada.
▶ 1:48:34And and then, you know, we've been talking about democracy prohibiting the use of AI during an election to create political messaging that contains deep fakes of candidates for office in New Hampshire. And so this um you know, we've been talking about federalism, but sometimes the states have to act. I also have some concerns.
▶ 1:48:58I fully agree with the chairman and a lot of the sentiment here that Congress does need to come together in a bipartisan way, but I think some of the AI companies are want this preeemption because they know that they can muck up the congressional situation, which isn't that hard to do, you know, and create the enaction so they can do whatever they want to do for as long as they possibly can
▶ 1:49:28do it. So with that long Professor Richards, the federal government often regulates in particular areas that affect interstate commerce, like air travel. The states have areas where they traditionally take the lead, like insurance.
▶ 1:49:47when it comes to the states making laws that affect AI deployment, what sectors or use cases should continue um to be within the state's purview where the go federal government um shouldn't get in the way.
▶ 1:50:04Thank you for your question, Representative Ross. think there are a numerous uh list of them and I I would I would hope that the the committee would indulge me that if I if I forget one uh there are too many to count. I would say uh in healthcare in the provision of medicine um the I uh I work a lot with our with our physician scientists at Washington University uh through the Cordell Institute and they are concerned about having access to AI technologies to treat their patients but also to be sure that the delivery of those treatments and the development of those treatments
▶ 1:50:34is done in a way which is consistent with the ethical and sustainably ethical practice of medicine. Um I mentioned in my opening remarks the the problem we have in the courts of hallucinated citations that states should be able to safeguard the integrity of their judicial systems and litigation processes by AI specific laws that general laws will not be enough in these cases given the particular uh affordances and applications that AI produces. I think about education um I believe it was Mr.
▶ 1:51:03Kylie that spoke about a moment ago. Um, AI does have a potential to help people in education. Um, but it also does tend to create massive plagiarism problems. So, and I'm being indicated to wrap up by the chair. So, I will uh pause there.
▶ 1:51:16Thank you very much. And I yield back.
▶ 1:51:18Professor, you uh are knowledgeable and we appreciate uh and that's why I didn't didn't stop it at the bell by any means. I wanted you to finish what you were working on. With that, we go to the gentleman from Wisconsin for five Thank you, Chairman. Uh Mr. Their uh you've written previously about California taking a um European style approach to regulation.
▶ 1:51:46Uh Chairman Jim Jordan of the full judiciary committee and Mr. Kylie and I were just in Europe last month talking with uh businesses um both European businesses and then uh American businesses with headquarters in Europe now within the EU many of them in uh Dublin actually uh and what they told us was that this type of exant regulation
▶ 1:52:16where anti-competitive practices are regulated before they exist It typically undermines or o or overburdens companies before they can scale and and it kills kind of a lot of the small businesses uh before they're up and running. And it's exactly what Europe has has uh it's why they have no gatekeepers. We're the gatekeepers.
▶ 1:52:42America innovates, China duplicates, and then the EU regulates. That's kind of where we're at right now on a grand scale. So if the US were to follow the EU's model of overregulating AI before it's understanding the risks, what impact would that have on AI development and competition?
▶ 1:53:05Uh because I believe that the EU is trying to create a space for themselves just like they are with the seven American corporations uh for the most part created in California. Uh there's one other one called Bite Dance. you might have heard of it. Um, but you know, now we have the EU telling us with the DMA, the Digital Markets Act, how we can function and how we can uh, you know, advance ourselves as as an American economy.
▶ 1:53:35It's it's very
▶ 1:53:37Yeah, you've got it exactly right, Congressman. Let's actually put some numbers on this. Um, I often when I'm uh lecturing to students any leading global digital technology innovators that are headquartered in the European Union today and I'm usually met with silence. There are a couple, but actually 18 of the 25 largest digital technology companies in the world by market cap are American-based companies. Only two are European. Most people can't name them. When I ask that question, most people say companies that are now defunct like Skype, right,
▶ 1:54:04and others. Innovation has died in the European Union. They've committed essentially continental continentwide uh technological suicide with a a regulatory model that is based on a sort of guilty until proven innocent mindset where every single technology or innovation is somehow nefarious and must be bottled up and preemptively uh regulated.
▶ 1:54:24This is why compared to the past where the United States and Europe were very very evenly situated in the early 90s, we went down two very different paths and our path are more pro- innovation, progrowth paths that really the Clinton Gore administration unlocked with a Republican Congress in a bipartisan way that yielded incredible benefits for our nation which made us the global leader. The household names in digital technology in the European Union today are American companies. What is America uh what is European Union exporting on digital technology front? Red tape. That's about all I've got
▶ 1:54:54Mr. Fraser, what would be some of the appropriate regulations that states could do a good job on? Uh, and and then how would we fold that into kind of at the federal level having some some oversight? What what's your thoughts on
▶ 1:55:11Yeah, thank you very much for the question, Congressman. Uh, I think that dividing authority on the AI development versus AI use threshold is very important. If states want to regulate the use of AI, the application of AI in schools for example, in health care situations, for example, those are instances in which states can truly run experiments because they are finite. They are in within their own borders and they're specific to their residents.
▶ 1:55:37But when we see states begin to enact proposals that are going to impact how AI models are trained and developed, that's necessarily going to bleed into other states raising profound extraterality concerns. So I think that a moratorum at the uh in Congress focusing on the difference between AI use and application versus AI development is a very helpful place to begin.
▶ 1:55:58What I would also encourage Congress to consider is the creation of a cause of action that allows non-state residents more means to contest the extritality extr territoriality of different state uh AI regulations so that we're not just waiting uh for California to regulate and just hope no one challenges it but empowering Americans to say AI is too essential to allow one big state to set the terms for the rest of us.
▶ 1:56:25Thank you for that answer. I um chairman before I yield back I just want to make the comment that I mean one of the concerns on many different fronts is how do you strike this balance between state development and and economies and and not not seeing an overreach like we have seen here in DC many times. So with that I yield back.
▶ 1:56:45I thank the gentleman gentleman yields back. We now go to yet another gentleman from California.
▶ 1:56:52Uh thank you Mr. Chairman. I'm a recovering computer science major and when I was studying computer science, I thought, "Oh, neuronet networks, they're never going to work." So, just take whatever I say with a grain of salt. All right. I'd like to um just note for the record what happened. Congress established in the House of Representatives a bipartisan AI task force.
▶ 1:57:13I was the co-chair and there were 12 Democrats, 12 Republicans and we all agreed on over 80 recommendations in a bipartisan manner, a number of which could be turned into legislation and instead the Trump administration basically says no we don't want Congress doing anything and we go to the states and not have them do anything when have zero regulation and the Trump administration tried to put in a 10-year moratorium ban on states that was opposed by 17 Republican governors 20 Republican
▶ 1:57:44attorney generals and 130 Republican state lawmakers. And then that 10-year proposed ban failed 99 to1 in the US Senate. A spectacular rejection of what the administration was trying to do. So now we're in a place where the actual reality is it's not whether we're going to regulate AI. It's do you want 17 states doing it or do you want Congress to do it? And with that lead in Mr. here. I um I know you were supportive of a mortorium.
▶ 1:58:13Your approach failed. So now we're in this new sort of uh position since it's failed. I'm curious what areas do you think Congress uh should regulate in because it's clear we're not going to preempt with nothing.
▶ 1:58:30What are the things that you think that would be helpful for the uh American
▶ 1:58:36Sure. Well, first of all, Congressman, I want to thank you for your leadership on this with Representative Olenalty, with the House AI task force, and then also the legislation that you did mention that you sponsored on this. I think that was a good starting uh building block for what we can do. We've heard many other members here today talk about the sort of things that NIS could be doing or the new Cassie, which again is a carryover from the AI safety institute.
▶ 1:58:56We could take some of the ideas that are already been per percolating at the state level including in California and New York and others to basically build upon what can be done in federal legislation and you can combine that with other sort of targeted actions. I want to remind everyone here people say congress doesn't do anything. Does everybody already forgotten about the take it down act passed overwhelmingly right? We can take targeted approaches to this and we can take broad approaches.
▶ 1:59:20The point is is that we can't have the technocratic design of regulation being done in a patchwork like this or that is going to create serious problems for American innovators as we continue to try to race against China to build out our capacity. We have to balance safety and innovation at the same time. So, we do need to have some preeemption. In my testimony, I spelled it out in quite detail how to do this, but reserve certain powers to the states.
▶ 1:59:41I want to agree with the Democratic attorney general of Massachusetts who said, quote, "Existing state consumer protection, anti-discrimination, data security laws still apply to emerging technology, including AI systems, as they would in any other context." That's exactly right. States can continue to do that, but we need to have a federal framework to make sure we get this done right. Thank you. Uh, just as an aside, I think there may have been um some disparagement with California.
▶ 2:00:05I just want to note Apple is headquartered in California, Google, Meta, Anthropic, Nvidia. Turns out that California does pretty darn well with the laws that we have. Uh, so uh, Professor Richards, I have a question for you. California is now proposing SB53. Have you looked at that in the California legislature at
▶ 2:00:27No, not not at the level of detail that I want to answer questions under oath on
▶ 2:00:30Okay, then that's fine. Now you in your testimony thinks that
▶ 2:00:34the gentleman before you came in Mr. Frasier actually has studied it and quite favorable in many areas of it. If
▶ 2:00:41Oh, so tell me about SB53. What is your view of it?
▶ 2:00:44I I earlier in my remarks I said that SB53 is the least bad state bill I've seen with respect to AI development. I think that it gets right a lot of the emphasis on information sharing that we know is essential to leading to better AI policy. So the sorts of disclosures that SB53 calls for from labs, I think is a very positive step. Uh I would like to see it done at a federal level and not at the state level.
▶ 2:01:08I also think that the whistleblower protections called for in SB53 are important to again contribute more information sharing, but I will note that for example, Senator Grassly has a whistleblower bill pending before Congress that I would prefer to be the vehicle for those sorts of protections.
▶ 2:01:24Thank you. I appreciate that. So, Professor Richard, let me go back to you. Uh, so your view is there should be no preeemption whatsoever. So, let me just sort of ask you this question you can answer because my time will be up soon. When a large model comes out, it goes through this enormous amount of training and then postraining and and all this and you have a model. So, let's say one state says, you know, we're going to mandate testing. Another state says, we're not going to mandate testing.
▶ 2:01:49And the third state says not only going to mandate testing, we're going to mandate the 27 specific areas where you have to test. And then another state says we're going to go even further than that and do you know 35 specific areas and be very specific what you have to disclose and on and on. Um how does even technically an AI company deal with that when they have one model? Do they just say we're just not going to be able to allow this to happen for example in Missouri or California or Florida?
▶ 2:02:18like how does it even work if you have 17 states regulating one AI
▶ 2:02:23And I I'll yield back and let him
▶ 2:02:26I I was giving you all that extra time so you can let him answer and if there's a followup within reason, we'll let you have it. It's the advantage of being nearly at the end.
▶ 2:02:35There you go. Thank you. Under that hypothetical, congressman, I think it would be very challenging for a company to apply. But it is not my position that there should be no preeemption, just that we should not consider broad preeemption of state AI laws. Under appropriate circumstances, a a sensible federal law would be naturally preemptive. And I would welcome uh a reasonable federal AI statute, just as I have welcomed and advocated for a a reasonable federal privacy statute, which the United States is the only advanced economy that has does not have
▶ 2:03:05Great. Thank you.
▶ 2:03:07And we now go right to the gentleman from California. Uh what does the hat say, Eric? Jimmy Kimmel Live. Of course, the gentleman is recognized for five
▶ 2:03:20Thank you, chairman. And I'll get to AI in a moment, but I'm not going to miss the opportunity to ask my colleagues, the proponents of free speech across the aisle, and the champions who sit with me to condemn in the harshest terms what is happening right now from our administration.
▶ 2:03:40The second late night comedian has been taken off the air because the president did not like a I want to first condemn in the harshest tones the murder of Charlie Kirk. He should be with his family right now. He should be with his children. He should be alive. He had a right to say what he wanted to say to who he wanted to say without any physical violence being brought his way.
▶ 2:04:09And Jimmy Kimmel had a right to say what he said, which didn't in any way suggest that somebody in the MAGA world had been responsible for the murder of Charlie Kirk. He was just pointing out what folks online were doing as Twitter detectives before any investigation had been completed were saying about the assassination of Charlie Kirk.
▶ 2:04:34And then he pointed out that Donald Trump, who did not go to Kirk's memorial service at the Kennedy Center over the weekend when he was asked how he was feeling about the assassination, did not address it, but rather went right to a construction project. And for that, Jimmy Kimmel was taken off the air. That's not who we That's what it looks like in China. That's what it looks like in Russia.
▶ 2:05:01That can't be what it looks like in And the foundation of this, the genesis of this was the president's FCC chairperson, Brandon Carr, sending a tweet that said essentially and giving an interview to a podcaster where he said essentially, "If ABC doesn't want to do this the easy way and suspend Jimmy Kimmel, we will do it the hard way and it would be government
▶ 2:05:31censorship." Maybe I was not loud enough in the past when Republicans spoke out against government censorship. And if that's the case, I will go back and revisit whether I could have been louder. But that does not mean that today Republicans are silent just in an effort to own the libs.
▶ 2:05:52If you didn't like cancel culture when you thought it was happening in prior administrations, you certainly can't look at what just happened in our country and accept that that is something we should live with and we should tolerate. So, I want to make it clear, there's going to be a Democratic majority in just over a year.
▶ 2:06:09and to the FCC chairperson and anyone involved in these dirty deals, get a lawyer and save your records because you're going to be in this room and you're going to be answering questions about the deals that you struck and who benefited and what the cost was to the American people because that happened.
▶ 2:06:30I want to now move chairman and I appreciate you holding this hearing to AI and ask our witnesses first and I'll start with professor Richards. Professor Richards, what is the risk to the country particularly to children if the government does absolutely zero on AI?
▶ 2:06:58as far as legislation, as far as what they see, privacy that's taken, biases that are reinforced, what do you see the risk could be? Thank you, Congressman. There there are a num a number of risks. Some of them are known and some of them are unknown, which I think is why it's essential to preserve regulatory flexibility by the states as well as the federal government to deal with these questions.
▶ 2:07:20Um we've already discussed um and uh uh at at some length um but I perhaps we can't discuss enough the the the losses that the parents who are succeeded behind me have suffered. Um when we have the in some cases the the the reckless or the rash deployment of software agents uh in children's lives. There was discussions about telling them about Santa but they've done much much worse. That's one of the risk exacerbating the mental health epidemic.
▶ 2:07:49There are risks to children in schools. Uh children don't read books anymore um because of AI models. States should be able to to address that pedagogically um with particular consequences for our critical thinking skills that are necessary for our democracy.
▶ 2:08:05Thank you.
▶ 2:08:06And also uh Mr. The I just want to coming from another meeting. I want to thank you for your remarks earlier about the FCC chair and his hypocrisy about censorship. With that, chair, I'll yield
▶ 2:08:18Would the gentleman yield for a second? as often happens, there's a nuance of total agreement here. And I just want to speak well of your championing free speech. uh and perhaps uh those who leave uh broadcast like our mutual friend Bill Maher uh might find an even greater uh place an even even greater amount.
▶ 2:08:43I do agree with you that we need to continue to promote free speech and your kind words related both in defense uh of one but also in uh on Charlie Kirk is very much appreciated and I look forward to continue to work with you. You and I have worked on a lot of issues and I think this is one we can work on as well. Thank you.
▶ 2:09:01Thank you. Appreciate it.
▶ 2:09:03And that only leaves me my job here is not just to ask questions but perhaps to try to close uh on as positive a note as I can of what we seem to agree on. And I'm going to use a kind of a question comment uh combination. I'd only ask that if I'm accurate you agree that I'm somewhat accurate as briefly as possible. Uh, and I'll start primarily with Mr. Frasier, but I want to make sure I have total agreement.
▶ 2:09:31Uh, if Congress offers laws and does it normally not by definition, unless we expressly trample on common law, common laws remains a tool of the states. Is that correct, Mr.
▶ 2:09:47That is correct. Absent very clear language. Yes.
▶ 2:09:50Okay. So that's one of our challenges to make sure that any preeemption is does not challenge existing preempt existing laws. In the case of uh uh of you will uh existing laws in states for example product uh liability laws. We've never preempted those even though we do have some federal laws.
▶ 2:10:17The reality is an unsafe product, a product that injures people, have a myriad of state laws uh that already affect it. For example, when we went from a man striking someone to a man on a horse striking someone to a man in a car striking someone, we didn't necessarily have to make major changes in the law. They all fell and they were none of them were federally
▶ 2:10:42Correct. And there's a reason why law professors laugh at the idea of the law of the horse. Law of the horse. Exactly. So, uh, Professor Richards, uh, you gave us a great deal of caution. Is it fair to say that if we clearly carve around any question of of common law preeemption and at the same time do not stop causes of action which are although perhaps automated by a bot and the like, still in fact follow that
▶ 2:11:13that that horse example. that in fact for the most part don't we meet the requirement of allowing the states to continue to protect their citizens uh as they have for 250 years?
▶ 2:11:25I I believe uh uh chair Isa that states should have the ability to continue to experiment um with with their own laws in addition to the common law.
▶ 2:11:36I and I I fully agree with you but uh and I'll go to Mr. Thy because this is this is both law and and policy. You mentioned number of times um ingestion versus output uh ingestion uh which cannot easily be done 50 different states and 210 different countries around the world.
▶ 2:12:02Isn't that also a a case in which the federal government must both lead on where the standards are particularly as to patent, copyright, other intellectual property and in order to have a single voice speaking around the world to other
▶ 2:12:18Yeah, that's right, Mr. Chairman. And let's be clear, like we wouldn't be here suggesting that like we should have 50 FDAs for food and drug standards or 50 FAAs for different aviation standards by every state such that planes had to change every state. That would be crazy, right? We don't want that model for AI either. We don't want the
▶ 2:12:36but a death by fentanyl, every state has a right to have charges in addition on the use side. And and let's just be clear. Let's just check off the the the generally applicable laws that would be exempt from either moratorium or preeemption. civil rights law and discrimination, unfair and deceptive practices and anti-fraud, competition policy laws at the state level, other consumer protection.
▶ 2:12:55In fact, the LAN act actually helps the
▶ 2:12:58We can go on down this list and then we can get into the lawsuits. The one thing America doesn't lack is an active trial bar, right? There's going to be a lot of ongoing lawsuits and we should throw the book at bad actors. There's always going to be bad actors regardless of technology. We have the capability to go after them.
▶ 2:13:13Dr. Bray, I don't want to leave you out of this. isn't one of the greatest cautions we heard today that we in fact have to make sure that when harm is done to anyone in a given state that they have a reasonable cause of action if it doesn't exist federally it must be available to them by in the states. Is that correct?
▶ 2:13:33That is fully correct chairman. Thank
▶ 2:13:36Okay. Uh Mr. Fraser, I'm going to uh sort of guide this another way. From a standpoint of federal laws, it's fair to say that for all practical purposes, patent, trademark, copyright, these are bastions of federal law because under the recognition that they all travel instate, they have to have one standard, rules of the road. Correct.
▶ 2:14:03It was very apparent to the founders that they did not want a patchwork approach to copyright and patent law. Correct.
▶ 2:14:09I think they also said that no no state could erect a uh uh basically a drawbridge and charge a toll to pass from one state to the other. They specifically understood that that that states might do that
▶ 2:14:22and it's a lesson we've learned throughout history with respect to uh for example attempts to change the length of a truck before it enters another state by Iowa was declared unconstitutional. We've been here before. We don't want a patchwork when it comes to national goods.
▶ 2:14:36Okay. Well, I'm going to uh I'm not not far exceed because I think we have the agreement that helps us in the guidelines. Um certainly um in the case of one that was mentioned briefly, Padra um which does deal with uh that with deep fakes with uh the digital likenesses and the like and which we do have bipartisan support and we look to to move.
▶ 2:15:03Uh that was an element uh today and I'd like any of you that want to uh comment further uh for the record to do I guess the the the last thing that we we all have to do is is recognize for the families that came here that we from this chair and I think you heard it from both sides of the aisle.
▶ 2:15:28We want to make sure that if we pass a law that further helps protect against the losses that you had that it consider exactly what happened in the case of your families. and that if we pass a law that in no way should it stop the causes of actions that may uh exist.
▶ 2:15:47Uh if anything uh we at a federal level for example a death by fentanyl we want to hold those who knowingly deceive and sell pills um purported to be some kind of a drug when in fact they are a deadly poison that they be able to be charged with murder as in some cases have been done at the state level.
▶ 2:16:08So, I can assure you from this standpoint, and I think the ranking member uh would not nod in any way, but yes, that that is a common goal and that we heard that message loud and clear. And I want to thank those who were here today um for their presence. Um and I and I want to recognize uh Mr. Johnson for something he wants to place in the
▶ 2:16:33Thank you. I have uh a couple of uh unanimous consent requests. I'd ask
▶ 2:16:38I know I'm going to like them.
▶ 2:16:40I'd ask unanimous consent to enter into the record a letter by Alejandro Montoya Boya the vice president for the center for civil rights and technology at the leadership conference on civil and human rights. a letter to you, Chairman Isa, and ranking member myself, uh, dated September 18th, 2025, which expresses the conference's views regarding the potential preeemption of states efforts.
▶ 2:17:06Without objection, so ordered.
▶ 2:17:08And also a uh to enter into the record a statement by Vered Hores, the chief of strategic AI partnerships at Bria AI, titled don't ban state AI laws. uh as well as a um letter from 17 Republican governors to Speaker Johnson and Majority Leader Thun dated June 27th 2025 opposing the AI moratorum in the big ugly bill.
▶ 2:17:40Without objection so ordered and in I too have unanimous consents. I ask unanimous consent that an extensive uh report and letter from the organization known as engine which to both of us uh which is a coalition of small startups been around since I think 2011 without objection will be ordered um an additional letter from the Americans for uh detailing the benefits uh
▶ 2:18:10versus the risk of 50 separate states that'll be placed in the record without objection. um a article on from Politico dated
▶ 2:18:27there we go dated yesterday and it's uh the California-W Washington tech fight heats up were placed in the record without objection additionally an political article uh from uh the entitled we don't want California to set the rules for AI across the country. Trump advisor says will also in spite of that being placed in the record.
▶ 2:18:54And I want to thank our witnesses. Uh you've been you've been informative. You've been helpful. And I think that this this has in fact uh furthered our understanding of quite frankly what our need to act and our need to act with a restraint from some of the warnings that were given by Professor Richards. And with that
▶ 2:19:26I have two more and then we're done. Well, I'd ask unanimous consent that the president's uh AI initiative be placed in the record in its full and additionally the the recent speech by Vice President Vance uh delivered in Europe be placed in the record without objection. Those both will be ordered.
▶ 2:19:58just to make it clear, additionally there will be general leave for similar items not specifically spoken to by members on both sides of the aisle. They'll have five days in which to submit those. And as such, we stand