“Building an AI-Ready America: Adopting AI at Work”

[Gavel sounds.] The subcommittee on Health, Employment, Labor, and Pensions will come to order. I note that a quorum is present. Without objection, the chair is authorized to call a recess at any time. Today's hearing will examine how artificial intelligence, or AI as we all know, is reshaping American workplaces and what that means for job creators and employees. AI is no longer science fiction. It is here and it is significantly transforming industries and workplaces. Optimistically, this technology may boost productivity, empower workers, and put the American dream within reach for more Americans. In my neighboring district, the Port of Savannah has invested more than $4.5 billion in infrastructure to support its supply chain operations, an investment that has made it the fastest growing port in the U.S. on the U.S. East and Gulf Coast. But as AI becomes more common in the workplace and fuels American innovation, we must ensure that workers are protected as well. To build an AI-ready America, we must first understand AI technologies themselves, how employers are using these tools, and the potential risk that AI adoption poses to workers. We want to help American workers and job creators compete and succeed in an AI-driven economy. Today's hearing is a strong step in that direction. Today we will discuss how the National Labor Relations Act can continue to protect workers' rights and promote stable relationships between employees and management in the age of AI. As AI adoption grows, labor laws and their enforcement should strike the right balance between protecting workers and allowing innovation to spur job creation and business growth. This hearing will also discuss the critical role of data collection to gain understanding of AI's impact on workers. The Bureau of Labor Statistics, BLS, plays an important role in tracking workforce trends, and it can help inform how AI is changing the modern workforce. Leveraging BLS insights will help us identify the challenges ahead and allow us to make evidence-based policy decisions. While technology can make our workplaces better, trust and transparency always benefits the workplace. Employers protect rights, strengthen morale, and build trust when they communicate openly about how AI is deployed in the workplace. That trust is essential for a competitive workforce and strengthens the ties between workers and job creators. The bottom line is that AI can create a lot of opportunities without compromising fairness or safety. We appreciate the insights our witnesses will offer about how we can ensure that innovation serves people and not the other way around. With that, I yield to the ranking member for an opening statement.

Thank you, Mr. Chairman. And I must say I could have written that opening statement by you myself. So this is a good sign. I think it's important for this subcommittee and this committee in particular because so much of the jurisdiction is here. I'm reminded as a member from the Bay Area, the fifth wealthiest district in the United States with a lot of investors in it, a friend of mine who was an early equity investor in social media called me a year or so ago and said they're doing the same thing they did with social media, meaning private equity investors as they're getting investment, but it's not all going to the primary drive is to get return on investment, which is nothing against investors, it's just we have to have it proportionate. And historically, the American workforce has not proportionately benefited from new technology. So we have to make sure, and I'm really encouraged by your comments, Mr. Chairman, that we approach it that way. And this committee and Congress is really responsible for this. And I will say as a former chair of the Senate Labor Committee in California, worked closely with the California Labor Fed that's working with responsible investors and employers to make sure we get it right. And again, as a member from the San Francisco Bay Area where a lot of this is happening, we want to make sure it gets it right. And also with three national laboratories and research facilities, both public and private, it's important that all of that is done the right way so we don't repeat the mistakes when it comes to technology that history has taught us that we should avoid with all the benefits from technology. So I want to thank the witnesses for being here. Again, I want to thank the chair. Artificial intelligence, or AI, has rapidly been introduced into Americans' lives. While some uses of this technology in the workplace show promise, it also poses significant dangers. Concerns about this new technology are not without merit. Some employers have used AI as a weapon in retaliation against workers who are trying to unionize or to suppress wages. The unrestricted use of AI can also infringe on workers' privacy. AI risks making it it harder for workers to prove it when their gender, race, age, or disability status are used to discriminate against them. As AI proliferates, we've begun to see employers, not all employers but some, use sophisticated automated technology to monitor and evaluate workers. This can lead to efficiencies, but we've got to value the balance here in workers' privacy rights. This includes tracking workers' on-the-job physical activity such as bathroom breaks, using facial recognition to assess their attention to computer screens and other tasks. That's pretty amazing. I think we'd all be concerned about that. We don't want to be monitored in this way, or at least we want guardrails and have the best available knowledge to first do no harm. In some cases, AI has the potential to make workers, consumers, and investors' lives easier and better. But without worker input, artificial intelligence could exacerbate the ever-widening power gap between workers and employers. AI should be used in the workplace to help workers do their jobs with ease and efficiency, not encroach on their privacy and pad the company's bottom line at the workers' expense. To strike the right balance, it is imperative that workers and their representatives have a seat at the table in the development and implementation of AI in America's workplace. And it is the role of local, state, and federal government to implement safeguards to prevent employer abuses and create this healthy balance. Congress has a responsibility to rein in big tech companies who take advantage of this prosperity and potential benefits so that AI can be used for everybody, workers instead of against them. We don't need another tool that will funnel profits to the ultra-wealthy at the expense of working Americans. Congress must take an active role in shaping how AI is integrated into our lives and developing thoughtful regulations that balance protecting workers with fostering economic growth. I will leave you all with a quote from a wonderful book that I've recently reread, The AI Con by Alex Hanna and Emily Bender. The authors warn that AI companies are selling consumers an attractive bill of goods that, quote, "a few major well-placed players are poised to accumulate significant wealth by extracting value from people's creative work, personal data, or labor and replacing quality services with artificial facsimiles," end quote. We have a responsibility and again, I want to thank my colleague and chair and our chair for setting the tone of this hearing. It's a balance. And I think we want to come to the table with the AI community, with investors, and make sure we strike this balance the right way so everyone can benefit from this and nobody gets taken advantage of from it. Thank you, Mr. Chairman. I yield back.

I thank the ranking member. Pursuant to committee rule 8C, all members who wish to insert written statements into the record may do so by submitting them to the committee clerk electronically in Microsoft Word format by 5:00 PM 14 days after this hearing. And without objection, the hearing record will remain open for 14 days to allow such statements and other extraneous material noted during the hearing to be submitted for the official hearing record. I note that some of our colleagues who are not permanent members of this subcommittee may be waving on to the hearing today for the purpose of asking questions. I will now turn to the introduction of our four distinguished witnesses. Our first witness is Mr. Bradford Kelley, a shareholder at Littler Mendelson P.C. from Washington, D.C. Our second witness is Ms. Revana Sharfuddin, a research fellow at the Mercatus Center of George Mason University in Arlington, Virginia, and I hope I got most of that correct. Our third witness is Mrs. Tanya Goldman, a fellow at the Workshop from Washington, D.C. Our last witness is Mr. David Walton, a partner at Fisher Phillips from Philadelphia, Pennsylvania. I thank our witnesses for being here today and we look forward to your testimony. Pursuant to committee rules, I would ask that you each limit your oral presentations to a three-minute summary of your written statement. As committee members have many questions for you, the clock will count down from three minutes. Pursuant to committee rule 8D and committee practice, however, we will not cut off your testimony until you reach the five-minute mark. I would also like to remind the witnesses to be aware of your responsibility to provide accurate information to this subcommittee. I will first recognize Mr. Kelley for your testimony.

Thank you for the opportunity to testify today on how we can build an AI-ready America. When I think about this topic, I think about my three-year-old daughter. Despite her frequent attempts to commandeer my phone, she doesn't yet know what AI is, but she will grow up in a world fundamentally shaped by it. As we consider the path forward, we must resist the urge to regulate in haste. Premature or poorly conceived laws can stifle innovation, create regulatory confusion, and give our global competitors an opening to outpace us in an area where American leadership is vital. AI is transforming the workplace before our eyes. Used responsibly, it can boost efficiency, reduce subjectivity, help prevent harassment, and enhance workplace safety. Critically, these benefits run both ways. AI empowers both employers and workers. Of course, like any technology, AI carries risk, including discrimination, intrusive surveillance, and job displacement. And while some might say that replacing lawyers is not the worst possible outcome, these risks deserve thoughtful, measured attention. But we must focus on genuine harms, not speculative fears. Many of the more extreme hypotheticals, like using AI to suppress union activity, remain almost entirely theoretical. As a former official at the Equal Employment Opportunity Commission and the Department of Labor, as well as in private practice, I have never seen employers use AI in a nefarious way. To the contrary, most are focused on using AI to improve efficiency and compliance. That's why a rapid rush to enact new laws is not the answer. As a former Army infantry and intelligence officer in Iraq War veteran, I was trained to master and fortify the defenses you already have before building new. The same principle applies here. The United States already has a well-established, technology-neutral legal framework fully capable of addressing most AI-related misconduct. We've already seen what happens when legislation gets ahead of itself. A prime example is the Colorado AI Act. The governor urged the legislature to fix its flaws on the same day it was signed. Meanwhile, New York City's AI law has been widely criticized for being ineffective and no complaints have been filed. Overreach by federal agencies can also produce serious and unintended consequences. In 2022, for example, the General Counsel of the National Labor Relations Board proposed an amorphous framework that presumptively deemed most AI tools used in the workplace to be unlawful. More broadly, some agencies have issued guidance without attempting to meaningfully understand how AI is actually being used in the workplace. It's also important to highlight what organizations are already doing to ensure that AI is used responsibly and complies with the law. Many leading companies have embraced self-restraint or self-regulation. Countless organizations have adopted and published principles on AI, developed internal accountability measures, and implemented policies ensuring that a human is kept in the loop. Employers are not trying to hide the ball. They fully recognize that transparency and accountability are essential. Finally, the growing patchwork of state and local laws has produced an increasingly unworkable regulatory environment. Congress should consider establishing a clear national standard that streamlines compliance, provides certainty, and preempts conflicting requirements. A coherent federal approach will better protect workers, support responsible innovation, and help ensure that America continues to lead from the front. Thank you, and I look forward to your questions.

Thank you. And I'd like to recognize Ms. Sharfuddin. Thank you.

Good morning, subcommittee ranking member, members of the subcommittee. It is an honor to testify before you. My name is Revana Sharfuddin. I am a labor economist and research fellow at the Mercatus Center at George Mason University. My research focuses on artificial intelligence and labor markets. Today my testimony focuses on the federal government's capacity to measure AI's effect on American workers. The three key themes are: first, AI transforms work differently than previous technologies, and it is harder to track with current data infrastructure. Second, our federal statistical agencies lack the tools to measure this transformation. Third, Congress should direct BLS and Census to close these data gaps. Previous automation replaced entire job categories. ATMs reduced bank tellers, robots replaced assembly workers. These changes were visible in employment statistics. AI operates differently. It automates specific tasks while leaving others untouched. For example, as an empirical economist, I used to spend significant time writing computer codes for economic modeling. AI now handles much of that, freeing me to focus on research design, collaboration, and communication. The occupation economist still exists, but the work has changed. This is precisely what we need to measure. For instance, while AI may outperform radiologists at reading mammograms, it cannot handle the other 26 tasks required for patient care. In fact, a study by Stanford economists found that across 950 occupations, although AI could perform certain tasks in nearly all jobs, there was no occupation that AI could fully handle alone. Our labor statistics count jobs, not what people do within them. The current population survey includes no question about AI use or task changes. The GAO flagged the data gap in 2019, recommending that the DOL develop better methods to track automation's workforce effects. BLS agreed and has taken incremental steps. But the pace of AI adoption has accelerated dramatically. There are three achievable steps that are targeted and low-cost. Adding an AI supplement to the current population survey would capture how workers use AI and whether it has changed their tasks. Linking firm-level adoption data to worker outcomes through existing census programs would let us track real effects. And coordinated annual reports across the federal agencies would give this committee the comprehensive evidence base it needs. The decisions this committee makes about workforce policy will shape whether AI delivers broadly shared prosperity. But good policy requires good data, and we currently don't have it. Thank you. I welcome your questions.

Thank you. I now recognize Ms. Goldman.

Good morning. I want to thank Chairman Allen, Ranking Member DeSaulnier, and members of the Health, Employment, Labor, and Pensions Subcommittee for the opportunity to testify on the impacts of artificial intelligence for workers and implications for their rights and safety. I have over a decade of experience in senior positions within the federal government's worker protection agencies, including at the U.S. Department of Labor and the Equal Employment Opportunity Commission. During my time in the Biden administration, I advised on the creation of the blueprint for an AI Bill of Rights, as well as rules and guidance to support employers and workers with the use of AI in the workplace. I am currently a fellow with Workshop, an organization that leverages policy expertise and partnerships to advance policies that protect and expand workers' rights. The speed of AI development and adoption necessitates policy responses that ensure workers have a voice and a seat at the table and the power to shape the design and implementation of AI. Congress can ensure safe and responsible use of AI in the workplace, including by shoring up existing workplace protections and enforcement, enacting new guardrails, and preserving states' ability to protect their own residents. Substantial data and research show that concerns about AI are not hypothetical. Workers are already experiencing harms. Employers are using AI tools to collect data, determine wages, and hire workers. This includes the use of algorithms to set compensation for millions of workers, which can suppress their wages. This includes, as a Roosevelt report on nurses noted, that on the same day, at the same hour, in the same hospital, two different gig nurses can be paid different amounts by the same app. These tools often replicate and exacerbate systemic biases. These systems lack transparency, so workers often do not understand how decisions affecting their livelihoods are being made. This can leave them without meaningful recourse to challenge the decision or to correct inaccurate data that is being used. Employers also collect and rely on data to inform automated management systems. This technology allows employers to surveil workers in more exacting ways than a human boss ever could. 90 percent of U.S. employers in recent surveys reported using algorithmic management for their workers. These systems can control scheduling and task assignments, performance monitoring and evaluation, and disciplinary actions. Time-on-task measures can pressure workers to work constantly and make it difficult to take necessary breaks, which can penalize pregnant or disabled workers if they take a break just to use the restroom or to take some medicine. Workers also feel pressured to increase their pace to unsafe speeds, which can lead to injury. During the Biden administration, DOL cited Amazon for injuries employees experienced from unsafe processing speeds. Constant invasive surveillance can also discourage workers from engaging in legally protected activities. These harms are the product of choices, not an inevitability. Congress has a role to play here in addressing these harms and shaping our future. Many of our existing workplace laws already provide workers with baseline protections, but they must be enforced and strengthened. America's workers also need broader federal guardrails on the use of AI. Majorities of both parties favor more regulation to limit AI's potential negative impact on society. Guardrails should include transparency, which has come up several times already, disclosure so workers know when AI systems are being used to make decisions about their employment and they can correct any inaccuracies. Humans should be involved in significant employment decisions. Developers and employers must test systems for bias. Workers and their unions should have a role in these processes. Many states have already enacted laws addressing AI-related workplace harms. These represent important progress in establishing baseline protections for workers. Federal legislation should be a floor, not a ceiling. We must meet the dynamism required of our modern workplace while also establishing robust protections for workers across the country. Congress has a role to play in protecting workers' opportunities, safety, and economic security. It can make sure worker protection agencies are enforcing existing laws, strengthen those laws, and enact new ones. I welcome the opportunity to work with this committee on policies that will safeguard workers' rights and ensure the fair and equitable use of AI in the workplace.

Thank you. I'd like to lastly recognize Mr. Walton.

Thank you. Good morning, Chairman Allen, Ranking Member DeSaulnier, and distinguished members of the subcommittee. Thank you so much for this opportunity to be here. I'm honored to be here, especially joined by my wife and my daughter who gets to see her old man at work. I'm Dave Walton. I'm a partner at Fisher Phillips and the founder and co-chair of the firm's artificial intelligence practice group. I've been practicing labor and employment law now for over 30 years and have been advising employers nationwide on AI governance since early 2023. My testimony addresses three key questions: how are employers using AI and managing risks, will AI destroy jobs or create them, and what's AI's impact on workers' section 7 rights under the NLRA? First, as everybody knows here, adoption in AI has exploded. Companies are deploying AI across every function, screening resumes in minutes, detecting safety violations in real time, flagging compliance risks, recommending training and upskilling for employees, identifying promotions and transfer opportunities that employees should pursue, and even anticipating employee burnout. This list grows daily and trying to keep on top of all the developments in AI is a full-time job. On top of this, AI has incredible potential to improve workplaces, eliminating mundane tasks, making work more enjoyable, making factories safer. But with great power comes great risk. That's why many employers are building governance systems. Governance means being intentional about risk management. It's about building a process, following the process, and documenting your attempts to follow that process. These frameworks include cross-functional teams that involve many stakeholders in companies, including IT, legal, HR, and compliance, bias audit testing for discrimination, human-in-the-loop requirements for major decisions like hiring and promotions, training employees on how AI works, on where it fails, and very importantly, how to report any problems that they see with AI. These governance frameworks work. Self-regulation through robust internal controls outperforms blanket restrictions, especially in an emerging technology like AI. Second, the jobs question. Will AI destroy employment? History and economics suggest no. AI will likely increase total jobs, not decrease them. Through Jevons' paradox, when technology makes a resource more efficient, it lowers costs, expands demand so much that consumption increases. Every automation wave that we have had, especially since 1750, has followed this pattern. ATMs reduced tellers per branch, yet total teller employment rose as banks opened more branches. Spreadsheets eliminated bookkeeping departments, yet office employment exploded. AI will likely follow suit. Some argue we need regulation to slow AI adoption. But AI is inevitable. Trying to stop AI is like standing in front of a tidal wave with a wiffle ball bat. It's just not going to work. The marketplace will force companies to embrace it. Companies that resist won't survive. Countries that slow adoption will fall behind. Third, employee surveillance concerns. Section 7 of the NLRA already protects organizing rights, and AI monitoring must respect them. The critical distinction: surveillance targeting union supporters violates the law. It already violates the law. Monitoring for legitimate purposes, safety, anti-discrimination, and productivity, is lawful when done with transparency, manager training on protected activity, technical safeguards, and worker involvement. Worker buy-in is critical. In sum, AI has become essential for American competitiveness. The employers I work with prove daily that robust self-governance protects workers while capturing AI's benefits. We can harness this technology responsibly through smart governance, not blanket prohibitions. I welcome your questions. Thank you.

Again, I thank the witnesses and appreciate your testimony. Under Committee Rule 9, we will now question witnesses under the five-minute rule. I will first recognize myself for five minutes. You know, in preparing for this hearing, we were discussing how our country evolved and how the workplace evolved. Obviously, it went from horse and buggy to automobile. That was a huge transition. And of course, prior to really World War II, most folks like me grew up on the farm and were employed on the farm, over 90 percent of the people of the country. Then we had the Industrial Revolution, and then of course the tech revolution, and of course now we have AI. So, and it's been amazing what the economy, how resilient the economy has been and how the workforce has adapted to every one of those stages in our country. Ms. Sharfuddin, you point out in your written testimony that federal labor market statistics were largely designed to count jobs rather than to measure how work is done. Can you explain how this creates a blind spot for policymakers as we consider AI in the workplace?

I appreciate that question. So currently, we ask workers what kind of jobs they're in, their wages and so on. But we know from literature, especially from theory, that AI is affecting jobs at a task level, and we don't have questions related to task level. For example, if there is, for example, an image designer or a marketing person at a firm, if you ask that person what jobs do you do, and that person can answer that they are doing design job, that is there in BLS. But what we need to ask is that do you use AI in your job first? And second question, what parts of your job is being used via AI? And then perhaps third and fourth question of how much of it is being automated, how much of it is augmenting, and how much are you learning via those tools?

Good. Thank you. Mr. Walton, you say in your written testimony that it is critical for employers who adopt AI tools to be fully transparent with their employees and to get their buy-in. Will you explain why it is important from an employer's perspective to get employee buy-in?

Yes, great question. If employees don't buy in to what you're doing, they are going to try to avoid it and they are going to try to work around it. We have seen that, like for example, there are some tools out there that were used by some employers that would measure if your mouse was moving and stuff like that. Well, there is now tools that you can buy that move your mouse automatically for you, you know, if you're monitored like that. So, in plus if you don't have employee buy-in, they won't trust you and they won't respect it. And so it's critical to be transparent with what you're doing for employees. Plus the other thing is it's hard to get good employees these days. So good employers are going to treat their employees well, and if not, then the market's going to speak and they're going to lose employees. So it's absolutely critical to get buy-in and to be transparent with your employees.

What are some of the most successful approaches you've seen employers take to get employee buy-in?

Well, I think you run pilot programs and you involve employees from the beginning, especially in the unionized work environment. You are up front with the employees and you explain to them why are we doing this? Why are we bringing AI in? What are we trying to accomplish? And it's also very important to add feedback loops where the people on the front lines, because it's the employees on the front lines who see all the problems. So it's critical to add feedback loops where they can see the problems and report back to management what the problems are with the tool.

Thank you. Mr. Kelley, I was struck by your written testimony that in your expertise and senior leadership roles at the Equal Employment Opportunity Commission and the Department of Labor, as well as in your private legal practice, you have never encountered an employer attempting to use AI to interfere with union activity or suppress union organizing efforts. Why do you think this?

Well, I think what Mr. Walton just said initially in terms of the fact that I believe most employers want to do the right thing. They realize that it is a symbiotic relationship, that it benefits both sides by doing the right thing. And I feel as if, you know, if you look at the laws that, you know, these historical laws that we have, the National Labor Relations Act is from 1935, the Fair Labor Standards Act from 1938, Title VII is from 1964. These laws were passed well before technology in the workplace, especially artificial intelligence, passed well before the internet as well. But these laws have been able to survive, and the current framework is really important. But I feel as if, you know, as far as employees specifically, why employers aren't looking at it that way is because of that idea of trust, that idea that you want to ensure that there is a good working relationship. During COVID, for example, a lot of employers turned to AI as part of the remote workforce. And a lot of times because they weren't able to do traditional in-person interviews, they were using AI tools. So it was increasingly important as part of that trade-off to make sure that they were being productive, which is why a lot of employers turned to AI tools at the time.

Well, thank you. Thank you very much. Now I recognize Mrs. McBath from Georgia for the purpose of questioning the witnesses.

Thank you, Mr. Chair. And thank you to our witnesses for being with us today. I have actually read all of your testimonies. There are fewer issues that stand to be as impactful to our society as AI, and it will be always that way in the near future. The American people are living through another Industrial Revolution, and Congress must do more to help Americans understand how they can make AI work for them but not against them. Atlanta is home to so many amazing artists and stagehands who bring our favorite music, shows, movies, and commercials to life, as you can tell I'm from Georgia. Whether they be writers, actors, painters, musicians, key grips, or any of the other professions that go into making Atlanta the Hollywood of the South, these jobs are incredibly important and they must be protected for the people who rely on them as their main source of income. Not only do they provide a good living, but these also aren't just any jobs. They are dream jobs for people who can't imagine themselves doing anything else. Jobs in the arts provide fulfillment that others can't. They are passions that people stick with despite being tough industries to break into. There is pride in being able to say that they stuck with their dream, that you owned, you know, you honed your passion into a skill that you not only truly love but really has value that other people are willing to pay for. People who built their careers and put out the best products are at risk of having their main source of income actually stolen from them. When an actor or singer has their God-given voice taken from them and used in content without their consent, that's theft. When an artist spends their life creating a specific style only for it to be plugged into an algorithm to create more content that they will never be paid for, that is theft. It's no different than stealing any other product that someone made. If anything, it can feel even more violating because what was stolen was your voice, your likeness, or your art, something truly that is only unique to you. Unions like IATSE, SAG-AFTRA, and the Writers Guild of America have already gotten major agreements and contracts to protect their workers. For example, IATSE and the Alliance of Motion Picture and Television Producers agreed that generative AI is covered union work, must be performed by a human, and that no crew member can be forced to input a prompt that would replace another crew member. This ensures that workers are given the opportunity to use their tools, not just be replaced by those tools. The right to join a union and collectively bargain is a concrete way for workers to have a real voice in their workplace. At a time when Americans feel like they have less power over their daily lives, unions give people a real say over the decisions that are being made that directly impact them and their communities and their families. Through their unions, workers negotiate better working conditions, higher wages, and fair policies on the job, like the use of or the implementation of AI. Ms. Goldman, can you briefly describe why unions are so uniquely positioned to help regulate the use of AI and protect workers from being negatively impacted and possibly even losing their jobs?

Yeah, thank you so much, Congresswoman. I really appreciate your comments and I agree that unions play such an important role in ensuring that workers have a collective voice in the workplace and that it serves as a counterweight to the improper, the imbalance of power in the workplace right now between workers and employers. And when workers have a voice and a seat at the table in the design and deployment of AI, as Mr. Walton said, their expertise can inform the technology and also enables the most effective interventions, which are won and implemented by unions through collective bargaining in how technologies are implemented in their workplace, including AI. You know, in addition to the examples that you raised, NewsGuild journalists have won protections in the newsroom from the use of AI, and this benefits them and it also benefits their employers, so we're not getting and consumers, so false and bad reporting isn't getting out, right. CWA has done a lot of negotiating around the monitoring of call center workers, and this really helps those workers make sure that AI is used for feedback and to support their performance but not used for excessive surveillance and discipline and evaluations.

Well, thank you. It's clear that unions are already playing an important role in being workers and helping workers to utilize the AI tools instead of being replaced by them. And I yield.

I thank the gentlelady for yielding. Now I call on our great Chairman Walberg for his line of questioning.

Well, thank you, Mr. Chairman. And thanks to the panel for being here. I almost, and I will hesitate, I won't do it, I wanted to ask if any of you have your prepared statements made by AI. I'd love someone to say yes and it fit exactly what I wanted to say. But thank you for being here. This is a key topic all over the world, and if we don't use it well, China will overtake us. And that's something we have to understand regardless of workplace, individuals, performing rights, and all of the rest. We have to be the leader. Mr. Kelley, your written testimony is critical of the way the Biden-era National Labor Relations Board approached AI. Will you explain the Biden General Counsel's approach and why you believe it was to be misguided?

Yes, absolutely. And it's an excellent question. So obviously there's been a lot of concerns about the use of AI in the workplace, which generated the General Counsel at the time, Jennifer Abruzzo, to issue a memo on Halloween of 2022. In this memo, it introduced an amorphous framework that created a presumption that employers using AI overall would be in violation of the NLRA. And it was flawed from the get-go. The entire scheme was based on a law review article or a law journal article from 2018, three sentences that included a footnote that said, "This proposal is laid out here only briefly to be elaborated on in future work," which was never released. And these three sentences in this 2018 law journal article established this framework. And I think that's also an important thing to kind of stress here is that a lot of the concerns that we hear about with artificial intelligence originate in academic literature. When we hear about all these problems about artificial intelligence, we're not seeing the cases. You know, as somebody who does a lot of litigation, the reality is that when there's a big legal problem, you see cases generating on it. You're not seeing the cases come up. All these problems that you hear about with artificial intelligence really are speculative in many ways and they can only be found in academic literature. But going back to the NLRB memo, it lacked clear definitions and standards, and it also failed to acknowledge how AI is actually being used in the workplace. The fact that employers are using it to prevent harassment, that they're using it to improve safety. It disregarded those and put them to the side. And also consider that the memo wasn't subject to notice and comment, there was no stakeholder feedback, there was no voice of employers, there was no vendor feedback, and those are the ones who are actually developing and using the technology.

Just a Halloween scare.

Yes, exactly. Boo.

Yeah, yeah. Thank you. Ms. Sharfuddin, your written testimony says that much of existing research suggests that AI is more likely to reshape tasks within occupations than to eliminate entire jobs. I certainly hope that that's true. Could you explain what in the research leads you to that conclusion?

Thanks, Chairman Walberg. That's a very important question. So right now, we don't see in our in 2025 labor data, we see that we're not seeing as many job created, but we're also not seeing that many jobs, you know, people that many people getting fired. So the unemployment rate is slowly getting up there, but also we have to make sure that the jobs losses we are seeing, are they in AI-exposed area? Are they in AI-exposed occupations? So currently, we don't have nationally representative data by federal agencies where we could study it very rigorously. All we have are some private data, for example, from Anthropic Economic Index. So that is privately done data from Anthropic, who's one of the largest AI technology company. But we have some issues with the type of people use that tool. So there is a selection bias in that tool because only certain types of software engineers and so on, they use the tool. So this is not nationally representative. So from those privately available information, we can see that the changes are happening at task level. Certain, for example, economist researchers, they are using it for editing and writing certain things, but collaboration, idea generation still staying within human worker. So we see it from private data, but we do need more comprehensive federal data from the whole country.

Thank you. Thank you. Mr. Walton, I'm sure your daughter's glad that your only dad joke that you used this morning was very fitting. I'll use it again, tidal wave, wiffle ball. Perfect. Talk to me about the imposed tax on employers that some are suggesting and why you think the robot tax is a bad way to protect American workers and what Congress should do instead in four seconds.

Yeah, in four seconds. I think it's a bad idea because AI is going to create more jobs than it displaces. I was reading something recently from McKinsey that the internet created 2.6 jobs for every job that it displaced. If you stop innovation in the American workplace, it's going to hurt the American worker and it's going to hurt the American economy vis-a-vis other countries like China.

Thank you. I yield back.

I thank our Chairman. Now I'll call on Mr. Mannion from New York for your line of questioning.

Thank you, Mr. Chair. Thank you to our witnesses for being here today. It is important that we're talking about responsible use of AI in the workplace because it's something that affects so many types of jobs and industries out there. And people understandably have different feelings towards it depending on their own experience. As the use of AI tools continues to expand, we must make sure that they're being deployed in ways that actually benefit American workers and benefit the American economy. We know the risks of AI as well as its utility. I certainly share the concerns raised by my colleagues, especially over worker privacy and job displacement. AI should never give employers a license to automate important decisions without human input and oversight. Nor should it be used for excessive control or surveillance over employees. We need common-sense policies that deter it from being used in these ways, and workers should always have a voice when it comes to AI implementation. We also can do more to promote the most beneficial applications of AI. Last month, I introduced the Immersive Technology for the American Workforce Act. This bipartisan bill would expand the use of immersive learning tools such as virtual reality and augmented reality to strengthen our nation's workforce development programs. These immersive technologies often integrate AI, and when combined, they hold enormous potential to improve job training and productivity across a wide range of industries. Ms. Goldman, my questions are for you. Obviously, with the adoption of any new technology, there are uncertainties about how the technology will be integrated into the workplace over the long term. Nonetheless, many employers are already experimenting with AI. Can you share what are the most pronounced sectors that are attempting to experiment with AI at this time?

Yes, Congressman. So we can see that it's very widespread. You know, I mentioned in my testimony that 90 percent of surveyed U.S. employers are using algorithmic management tools in the workplace. A survey done by my colleague, former colleague Alex Hertel-Fernandez of U.S. workers found that two-thirds of them reported that algorithmic management tools were being used in their workplace, and this was true across earnings in different jobs. In the recruitment and hiring space, we see kind of variable data, but a third to the vast majority of employers are using AI tools in hiring and recruitment. And this is a range of industries. I mean, it's really pronounced in warehousing, we're all very familiar with what warehousing workers are experiencing, rideshare drivers, delivery drivers, but we see this in lots of sectors. It's retail, it's healthcare, it's government, it's manufacturing, it's hospitality, it's call centers. So this is why it's really important that we have guardrails around this technology to protect these workers.

Is there a certain sector where experimenting in AI has actually shown that it has been very effective and also assured that jobs are protected?

I mean, I think this goes back to why we need workers to have a voice and seat at the table and be involved in the design and deployment of AI. So even in call centers where we know there's extensive use of AI tools, having that worker feedback is critical. You know, there are workers who've reported to us that they are prompted by the AI to give inaccurate answers to their customers. That doesn't help the customer or the employer or the worker, and they have to do it or they'll be suffer through their performance evaluations. But by being able to provide that feedback, they can work with the employer to make sure the AI is being used in a way that is productive for all involved and not harming the worker and causing excessive surveillance and monitoring.

Can you speak to AI and its role in the job application process and where there may be some shortcomings?

Yeah, so automated hiring systems are essentially acting as digital gatekeepers to workers' economic security, and this can result in discrimination. It can replicate and deepen existing inequities in our society. These algorithms can be trained on inaccurate or biased data. It could be unrepresentative. They could work from proxies that do not actually tell you how much a worker is able to perform the functions of the job. And they can pull from thousands of data points from social media, from the internet, from third-party data brokers. So it's not just, you know, the interview and application process that the worker is being evaluated on.

Great. Thank you so much for your testimony. Go ahead.

I have 12 seconds, but just to respond to something that Mr. Kelley said. You know, I just want to note that it is not a hypothetical concern that surveillance is affecting organizing in the workplace. Data collected by employers can definitely have a chilling effect and impede worker organizing.

This former union president thanks you. I yield back.

I thank the gentleman for yielding. Now I'll call on Mr. Owens from Utah for your line of questioning.

Thank you. Thank you, Mr. Chair. First of all, thank you for this, this conversation. It's very, very important. It's a game changer, obviously. I want to say is AI is an incredible tool, has potential to strengthen our workers and national productivity if used correctly. I look forward to the hearing of witnesses that talk about the promise of AI and where we can take it from here. I want to, Mr. Kelley, I want to kind of follow up on a question that the Chair asked, the Chairman had asked earlier. When it's talking about the report that came out in 2018 that has kind of tainted the way people use or looking at AI today. What can we do to make sure we're mitigating that fear factor? Because it's prevalent. We have to make sure that we are embracing AI, and the only way to do that is again through education and using it the right way. So I don't know if you want to add to that conversation you had earlier.

Yeah, I think that it's really important that when you look at a lot of the examples out there of these worst-case scenarios with artificial intelligence, a lot of them are very outdated. They're from 2018, they're from 2019. Think about how much technology changes in how much it improves. Think about like the latest update to your cell phone. You get your new cell phone, it's got a much better battery life, it's got incredible new features. You're like, wow, this is incredible. We're not recognizing that with artificial intelligence. And the problem is as we hear a lot from the employee side or the, you know, union side about, well, we need new guardrails, we need a seat at the table. What does that mean? What as far as what laws do we need in place? If you're using artificial intelligence to screen applicants on the basis of race or sex or disability, those violate existing laws. We need to put the focus on the existing laws. And I do think that even for these hypothetical type scenarios is that it's really important that, so I came from the EEOC. I was a senior official at the EEOC before joining Littler. And we launched the AI initiative in 2021 to look at how AI may result in discrimination in the workplace. And that initiative has been completely suspended. It's been because there hasn't been these cases. There hasn't been this employer activity that was reported on. I mean, you hear about these concerns and again they're completely found in academic literature. They're found in these hypotheticals. That's why I do academic articles, because I try to understand what the other side is saying on this. And the cases aren't there. And I think that's really important because the problem is by making these speculative concerns and these hypotheticals, you're going to hamper innovation. You're going to have employers who are going to say, well, let's not go full throttle with artificial intelligence, let's hold back. And that allows our adversaries overseas to get ahead of us. And I think that's a really important thing to keep in mind, and we cannot allow ourselves to get outpaced in a lot of ways. So when I hear about all these concerns with AI, the existing framework addresses those, and that's really important because we can't have laws that are passed that have no purpose. You know, as I mentioned before in my opening, Colorado AI Act went into effect and then on the very same day the Colorado Governor said, "This is a flawed law, we need to revise it." On the same day it was signed. A couple days later, the Colorado Governor and the State Senate Majority Leader and the Attorney General sent a letter to the business community saying, pledging to revise the law. Why are we doing that? And it doesn't make sense.

Thank you. Thanks so much. I want to make sure I have one of a couple of other questions here. Ms. Sharfuddin, you point out in your written testimony that small businesses are adopting AI a little bit slower. Why do you think that's happening?

Thank you for your question. Small businesses generally face higher fixed costs when they are adopting any new tool or adopting new technology or expand their business. So that's one of the reasons that they lack the capacity. But we are seeing that as AI is diffusing through the economy, we see that they are adopting it more. I think I would run out of time, but I would like to refer to a policy brief I wrote a few months ago about how we can empower small businesses to first adopt AI but also train their workers. And we can do it via tax expensing provisions that would empower 30 million, which is 80 percent of our economy, to adopt small business with their workers. Thank you.

Thank you. And I'd like to know more about that because small business obviously drives our economy in so many different ways, and we need to make sure that innovation is there for sure. Real quickly, Mr. Walton, you're familiar with AI discussions around AI in terms of being a job creator or destroyer. And your testimony provides insight on how it can retain employees and thereby reduce turnover. Can you kind of share a few thoughts on that on how that works?

Yes, sure. My testimony talks about for an example an approach by the University of Nebraska that was used to spot burnout by nurses. And so that's the and one of the things that they did is they stopped or they put a deep reduction in their nursing turnover. And so that is just one example of of how AI can be used to improve the workplace, to keep employees happy. AI is also being used to help identify for employees where the best transfer opportunities are in their companies, because it's better to have companies doing different jobs, especially in large companies. AI is also being used to help identify bad managers, you know, and so there are and these are just the tip of the iceberg. There's a thousand examples like this. I mean, the AI is much more of a net positive in the workplace than it is a net negative.

Thank you. Thanks so much. We need to bring innovation and heart to this conversation, and no place better than to do it in here. So thank you so much for your comments for sure. Appreciate it.

I thank the gentleman for yielding. Now I'll call on Mr. Scott, the ranking member of the full committee, for his line of questioning.

Thank you, Mr. Chairman. Several of witnesses have indicated that we might create more jobs than we lose with AI, but that still acknowledges we'll lose jobs. What kind of job training should we be developing to account for the displacement? Ms. Goldman?

Sure. I mean, I think, Congressman Scott, there are a lot of really important questions about workers' future with the implementation of AI in the workplace. And I think it's really important that we do the things that we know we need to do right now. We need to improve job quality so workers have access to fair pay and benefits. We need to make sure workers have access to unions, which involves updating our labor laws that are on the books. And then we need to, you know, make sure that workers, as we've been talking about, have a voice in the workplace. And this includes shoring up our existing social safety nets, including our job training systems, so that if workers do lose their job, they have a bridge to another high-quality position.

Anyone else want to talk about what kinds of job training initiatives?

Actually, that helps me tie to my previous comment. I think AI is moving so fast and it's really hard to think about federal or centralized way to see where the training is necessary because American economy is huge and it's dynamic and its drivers. So you have to think through policy, how can we empower businesses that they can train their workers and they can do they can response to the local labor demands. Because the labor demand, for example, in Colorado is going to look different from Washington, D.C. So in my tax policy proposal, I suggest that small businesses and businesses of all sizes are allowed to expense the cost that they put in for training their workers. And it would equalize the level playing field we have with technology right now. Businesses are allowed to expense their cost for when they buy a building or through bonus depreciation or when they buy a technology or research and development investment, they are able to expense it completely, but they cannot do the same when they train their ...workers. So I would recommend that we kind of equalize the human capital investment with capital investment in terms of tax expensing. Thank you.

You mean an addition, it's already deductible. You mean an additional expense?

Congressman, there are mazes of regulation for businesses to go through and I have a policy brief. I'm happy to send it to your office.

Okay. Thank you. Anyone else have any ideas on job training, what we ought to be doing?

I'll just say in general, Congressman, that AI has tremendous potential for education and that includes upskilling and reskilling employees. And I think it's very important for us to be thinking about ways that we can use AI to help upskill and retrain employees.

I think that we should look at innovative programs, sir. One that sticks out in my mind, especially as an Army veteran, as a combat veteran, is the SkillBridge internship program. And that's where at the last six months of your time in the military, you're able to do an unpaid internship with a private company. And it gives you the ability, the whole idea of building a bridge from the military to the civilian sector. And I think that looking at different ways, creative ways of having training because with artificial intelligence, these tools are incredibly developing at an incredibly rapid rate. So I think that just focusing on education overall and looking at new innovative ways to approach it will be very important moving forward.

Mr. Kelley, you acknowledged that discrimination could happen with AI. If the AI is used to screen applicants, you could have a discriminatory, you could have discrimination and that's illegal under present law. How would anybody find out, particularly in light of the fact that the EEOC is diminishing, is taking away your ability to see a disparate impact?

Well, I think that it's really important to say that as far as establishing that baseline, that in a lot of ways the use of artificial intelligence to screen out resumes or find the best applicant, it takes away that human element, which oftentimes is the most discriminatory part. So it increases the ability to make objective decisions. So you could take away the consideration of names that are oftentimes associated with, let's say...

Well, you can. I mean, if you've got discrimination going on, how do you deal with it?

The existing law. I don't think that there's any problem with the existing law. And I do think...

But if you don't have disparate impact, how are you going to prove your case?

Well, there's also the ability for commissioners to bring commissioner charges from the EEOC. That's been something that's been...

But the EEOC said they're not doing disparate impact. How would they even know what to target?

It doesn't have to be just disparate impact. I mean, it could also be disparate treatment. So they don't have to just focus on the disparate impact because one of the interesting questions with AI is at what point does disparate impact become disparate treatment? And there's kind of overlap, you know, with that area. But...

Thank you, Mr. Chairman.

I thank the gentleman for yielding and now I call on Dr. Onder from Missouri for your line of questioning.

Thank you, Mr. Chairman. Thank you to all the witnesses for appearing at this subcommittee today. I'm very interested to learn more about what AI might have to offer to our economy. I believe AI does have enormous potential. I mentioned at a previous hearing on AI, I am a physician and I know some of my colleagues now use AI scribes to decrease administrative burden and clinical documentation. I remember back in the day when before EHR, when I would bring home paper charts and spend hours dictating patient notes. Now AI can listen in and sometimes complete hours of paperwork in merely minutes. I have an Amazon distribution center in my district and I watched as AI was used for sorting packages, routing packages, extremely impressive. These are examples of just tremendous efficiencies of AI in the workplace. But despite the potential benefits, a recent survey showed only six percent of workers believe that AI will lead to more job opportunities for them in the long run. A recent survey found that most workers are worried about AI in the workplace. Mr. Walton, in your testimony today, you explained Jevons Paradox, which says that as technology increases the efficiency of how a resource is used, the total consumption of that resource will increase rather than decrease. If this holds true for AI adoption, the technology could actually create more jobs in the long run rather than take jobs away. Can you share examples of how AI efficiency might produce higher economic activity and if possible, more jobs?

Sure. I mean, we're seeing that already in the coding world. And we're seeing coders being able to do 10 times more the work that they've been able to do in the past. They've been able to do a lot more output. And if you go back to the internet, for example, there was a lot of fears when the internet was adopted that it's going to cost a lot of jobs. But just think of all the new jobs that the internet created from cloud computing, from website design, from e-commerce. I mean, I've got my daughter here today, social media influencers. I mean, there's whole new economies and whole new industries that were developed by the internet. Plus, I think, Congressman, AI is going to develop almost more like cottage industries because it's going to give power back to the people more. I mean, I have my phone, it's like having a Palantir in my pocket now with these types of AI tools. I can sit at my kitchen table and run a 10-person law firm just me. So that's going to give power back more to the people to create value for themselves. And so I think we're going to see a lot of that in the future.

Well, you know, in this discussion, I'm reminded of a famous story where the economist Milton Friedman was touring a quarry in China. And he noted that they were using picks and shovels. And he asked the foreman, well, why don't you use, in the United States, this would be run with tractors and excavators and backhoes. And they said, well, with picks and shovels, we employ more people. And Milton Friedman looked at him and said, then why don't you give them spoons? So Mr. Kelley, in your testimony, you mentioned the Colorado AI Act, which apparently was so poorly crafted that Colorado's Democratic governor had to tell his legislature to rewrite it. Could you walk us through some of the poorly drafted provisions and lessons that Congress might learn from that experience?

Yeah, I think that it's a prime example of just the rush. I think that they were trying to outpace the EU AI Act, the European Union's AI Act. And as part of that rush, they tried to really kind of force it through because they wanted to be seen as a leader among the states. So on the very day it was signed, the flaws were acknowledged and they pledged to the business community along with the state attorney general and the state senate majority leader, pledging that they would revise the law. But the problem is that it was supposed to go into effect earlier this year and it's been delayed till June. So a lot of the fundamental problems are the same. There's a lot of ambiguity with what the obligations are on developers and deployers. There's a lot of kind of as far as questions over how should this be interpreted. And so I think that overall, it's just very flawed and it will likely, it's quite possible that it's going to be delayed again. But I think that the New York City law is just as bad in a lot of ways. That law was rushed through, it received a lot of criticism from a wide variety of groups. And as of July of 2023, when it went into effect, it hasn't received one single complaint that we're aware of.

Thank you. I yield back.

I thank the gentleman for yielding and now we'll call on Ms. Bonamici from Oregon.

Thank you very much, Mr. Chairman, and thank you to the witnesses. Thank you for allowing me to join this subcommittee. I serve on the Education, Early Childhood, but also Higher Ed, but also on the Committee on Science, Space, and Technology and spent a long time on the bipartisan AI task force. So I'm really interested in the conversation today. Obviously, AI is reshaping the workforce faster than many employers and employees can respond. It's my understanding that about one in five workers is expected to see major job disruption and almost all workers are expected to see some of their tasks affected. What's concerning to me is that if this happens without proper education, security, privacy, and guardrails, it's really concerning. I'm actually working on a comprehensive human-centered framework for AI that will strengthen the teaching and learning, prepare workers, and address the security and privacy concerns because responsible AI policy should prepare people for a changing economy, not just blindly deploy new technologies. I'll add anecdotally that I was recently talking to someone who was grading college papers who said there's an extraordinary amount of time trying to figure out if it was written with AI in violation of that institution's rules. And students are really concerned because their professor decided to just give them a blue book test and they don't know how to write. So when we look at the 21st-century job skills that have been identified as critical thinking, creativity, collaboration, communication, we have to get this right. So Ms. Goldman, what specific actions can the federal government take to identify occupations most affected by AI deployment, following up on Representative Mannion's questions, and what resources or policies would be best to mitigate AI-related workforce disruptions?

Thank you, Congresswoman, and thank you for your leadership on these issues. And I think first we need to shore up our existing legal frameworks and make sure that we're enforcing the laws we do have on the books. As Congressman Scott was getting at, these are really black box systems that workers lack transparency. They don't know what tools are being used and therefore they can't challenge them. We need the EEOC conducting systemic and disparate impact investigations. And we need to shore up our labor laws, which haven't been amended in over half a century. We need new comprehensive laws. So I'm delighted about the work that you're doing. And these really need to ensure that automated employment and management systems are fair and are being regulated. These new laws can have things like transparency and disclosure, which almost everybody here agrees is extremely important for workers. They can have right to human review, make sure that we don't have robot bosses firing our workers.

Right. Thank you for mentioning my No Robot Bosses Act, which is bipartisan H.R. 6371. If anyone wants to sign on, we invite you. I'm leading this with Representative Moylan and DeSaulnier and we're also looking at reintroducing the Stop Spying Bosses Act, which is to protect workers from invasive surveillance.

Yeah, and that's so important. And then just last thing I'll say is we need impact assessments and I think Mr. Walton has talked about that as well and make sure that these tools are being tested for validity, reliability, and non-discrimination and that workers and unions should be involved in that process because they know how they're being implemented in the workplace.

Absolutely. And I heard the example of Colorado with the governor of Colorado, a former member of this committee. But I also have to say that until there's federal regulation, I mean, we spent a lot of time on the AI task force, bipartisan AI task force, but we didn't pass anything. We made recommendations, we issued a great report, but we didn't enact anything to protect against bias, to protect privacy and security rights. So until we do that, we need to respect the states that are acting to protect their states. So Ms. Goldman, what can Congress do to support employers who want to responsibly adopt AI in the workplace?

Yeah, and if I can just note, Congresswoman, how important it is that we do allow states to protect their own residents, including consumers, workers, and children, and make sure that they're able to respond to this rapidly evolving technology. And if Congress does pass a federal framework, which I hope it will do so, it can provide a baseline of those protections. But those protections need to be a floor and not a ceiling.

Right. And with regard to the existing laws, yes, they need to be enforced. But we had a lot of conversations about how do you enforce and where do you enforce? If it's a labor issue, does it go to labor? If it's an education issue, does it go to what's left of the Department of Education, assuming that there is a Department of Education? If it's a health issue, does it go to HHS? Is there some separate enforcement agency? So it's not as simple as enforce the existing law because it's complicated. So I just want to thank you so much for your testimony, all of you, and look forward to the ongoing conversation. I yield back. Thank you, Mr. Chairman.

I thank the gentlelady for yielding and Ms. Lee from Pennsylvania is recognized for five minutes.

Thank you, Mr. Chair. My Democratic colleagues have shared some of the ways adopting AI in the workplace can come at a cost to our civil rights and that of privacy. One would hope that officials at the highest levels of government are weighing these costs when adopting AI in their workplaces. But we have a White House posting AI-edited photos of civil rights attorneys and a president who's posting racist deepfake videos of Democratic leaders and AI-generated videos of himself dumping feces on Americans protesting his authoritarian policies. And before Elon Musk went back to X to profit from pornographic deepfakes, he led a government agency that relied heavily on AI. It's an agency that we now know stole Social Security data to try to overturn election results. And let's not forget the environmental toll of AI data centers. This administration is driving up electricity costs and straining our water supply to store data, spread disinformation, and silent dissent. Republicans are talking about responsible AI use while refusing to hold their own leadership accountable, just like they continuously refuse to hold corporations accountable. We must ensure that AI is not weaponized to undermine workers' rights, especially their right to organize. Mr. Kelley, just a yes or no, does your law firm, Littler Mendelson, use or encourage clients to use AI or other algorithmic tools for union avoidance strategies?

That issue has not come up.

So yes or no, would Littler Mendelson and its clients inform workers if AI were being used to persuade them against unionizing?

We advise employers on the use of artificial intelligence. We advise on specific, it's very...

That's a yes or no. It is yes, you would advise them?

There's no yes or no answer to that.

No, I think there is. But I'll move on. You claim that corporations' use of AI to suppress union activity is largely theoretical. And you said you've never seen employees, or sorry, employers use AI in a nefarious way. But it's hard for me to believe that one of the most notorious anti-union law firms in the country is unaware of nefarious practices. So like when Whole Foods, which is owned by a client of Littler Mendelson, Amazon, used AI to track which of its Whole Foods stores were at risk for union activities. Or when the National Eating Disorders Association fired all of its helpline workers and replaced them with an AI chatbot two weeks after workers voted to form a union. A chatbot they later had to suspend for giving harmful advice. We know corporations are not legally allowed to union bust or steal wages or discriminate or create dangerous working conditions just because AI helps them do it more easily or gives them more deniability. But worker protection laws are only effective if somebody is enforcing them. Ms. Goldman, are worker protection agencies like the Department of Labor and EEOC fully equipped to enforce worker protection laws in the age of AI and under this administration?

No, Congresswoman. I think as you know, they're chronically underfunded and understaffed. My colleagues at the National Partnership for Women and Families put out a report recently that the EEOC has 1,000 fewer investigators than it had in 1980. Yet as you know, they have 60 million more workers that they're trying to protect. They receive thousands more charges and they have more laws that they're enforcing right now. And the same holds true at the Wage and Hour Division. I think as of May 2025, they had about 600 investigators to protect the rights of 165 million workers in the workplace. And in addition, they need to be able to hire technologists and really understand the data so they can get into the black box and do the kind of systemic work that needs to happen.

Thank you. We can't replace federal worker protection agencies with corporations and tech monopolies with non-existent accountability policies. The last administration change showed us just how flimsy tech companies' standards are. The minute big tech platforms could step back from any accountability, most did because profits will always come before workers' rights and as long as Congress allows them to do so. This is exactly why we need legislation like the AI Civil Rights Act or my Eliminating Bias Act so that there are real guardrails and investment in place that demand transparency, accountability, and strong civil rights protections. Republicans are far too eager to dive headfirst into an authoritarian regime fueled by tech leaders. Workers, not corporations or union-busting law firms, are going to suffer if we don't implement real guardrails. I thank you for your time and I yield back.

I thank the gentlelady for yielding and now we'll call on the ranking member, Mr. DeSaulnier, for his questions.

Thank you, Mr. Chairman. And I want to go back to your opening comments and ours that I'm hopeful that we can work together on this. I feel like we're starting to fall back into a pattern where we're jumping to conclusions based on our historical perspectives. Ms. Goldman, I want to ask you about ghost jobs. I've asked this a lot, spent a lot of time on it. Mr. Kelley talked about, sorry Mr. Kelley, I'm putting words in your mouth, but this is the purview of us up here, is the wonder of algorithms. Mr. Walton, you alluded to this too. And I'm struck by the research on ghost jobs in particular and cognitive development on social media that's pretty common knowledge at this point. So ghost jobs research now tells us almost 50 percent of the jobs posting are not real. It's data collectors trying to get more information. And on the other side, our young people in America being creative, up to 60 percent of the applications are not real applications. So this goes into privacy and algorithms. How do we get control of this and be transparent so we make sure that people aren't unwittingly giving up more data and more data so the algorithms can be more predictive and more targeted? And in addition to that, all the cognitive development where CDC discovered that a third of adolescent girls in this country are being targeted specifically by YouTube's algorithms who are predisposed to depression to deliver them to YouTubes on how to commit suicide. And all of this was part of the algorithms according to the research at CDC to actually get young women in particular as they develop cognitively to be predisposed even further to depression. So those two things, how algorithms and AI are affecting the workforce, but also our young people.

Yeah, thank you, Congressman. And so let me tackle kind of both the collection of data and privacy in the workplace. So we have a lot of data brokers buying and selling information, including about workers. And when they sell that information to employers, employers can use it in ways that are potentially discriminatory, invade their privacy, or could be used for anti-union purposes. Employers can also sell information about their workers back to data brokers and that can be done without the workers' knowledge or consent. So we have an opaque market here that needs regulating and raises a lot of concerns about workers' privacy and civil rights. We know workers do not have sufficient privacy rights in the workplace. Employers are collecting information every day from them from electronic wearables, geo-tracking, facial recognition software, cameras in their cars, their trucks, their home computers, voice recognition software, keyboard logging. I could go on, but there is a lot of surveillance used to collect information, also the data that they're collecting from social media, the internet, and data brokers. And so I think we really need to consider making sure that data is not being collected if there's not a business purpose for that data, it's not being stored longer than needed, workers know what data is being collected about them, that data is safeguarded and won't be exposed to breaches, and that it is not being sold.

George Santayana said we're condemned, for those who forget history are condemned to repeat it. Having grown up in Lowell, Massachusetts, and listened to my grandparents and their generation talking about the Industrial Revolution, there was certainly great benefits that you alluded to. But workers didn't get protections because employees, employers gave it to them. I'm reminded of one of my heroes, who was the first woman cabinet secretary, Secretary of Labor Frances Perkins. I have an original photo of her and the Triangle Shirtwaist Factory fire where she as a young upper-class woman went across Washington Square and watched those young people fall to their deaths because we didn't have worker protection. So let's talk a little bit about how we got child labor laws during the Industrial Revolution for all those benefits. History repeating itself, employers just didn't, not all employers, just took care of their employees. So we should be aware of not being modern-day Luddites, but we should also learn from history that workers didn't benefit this without them responding to it. So the first labor union was the Amalgamated Textile Workers, which were young women who came from the farms and started losing their fingers because there wasn't workforce protection. So talk a little bit about how we responded and protected workers so we don't do the same thing here, that we see all the benefits of AI, but we don't have an appropriate protection for workers and not just trust people. We have to, to your comments, Mr. Chairman, we have to work together to make sure we don't repeat the mistakes of history.

Yeah, thank you, Congressman. And it is so important that workers and unions have that voice in their workplace and that collective action. As you know, the NLRA has not been updated since the late '50s. And so workers' interest in joining unions and being able to protect themselves in the workplace is surging, but private sector union density is at historic lows. So yes, as Mr. Kelley says, we have laws on the books, but we need to update those laws and enforce them to make sure that all workers have access to them. And then of course, it is urgent that Congress act. It is long overdue that we have additional protections for workers that respond specifically to the challenges that they are facing from AI in the workplace.

Thank you.

I thank the gentleman for yielding. Votes have been called, so we're going to try to wrap this up. So I will now call and recognize the ranking member for his closing remarks and I'm going to make mine quick and I would like you to so we can get to the votes. [Laughter.] Subtlely is one of your real strengths, as is mine. I just want to thank you for having this hearing. I think we could do really good work here, Mr. Chairman, together and it would benefit the employers and the employees and learn from history. With all due respect, history teaches us that workers don't benefit at the same rate of investors. So let's get ahead of that. Let's make sure that we are not just assuming the marketing that this is going to benefit everyone. So I'd be thrilled to work with you and all the panelists to have an honest and sometimes honest means sometimes we're not going to agree on this to get to a conclusion with all the institutions that are sort of set up on both sides to go to warfare on this. We reserve those, that's my expression, but let's work ahead of it and see if we can get it right. And to the EU, I think we can learn from them too. They're doing some really good things and I think the states are responding to that. I know California is as well. So if you're worried about us getting it wrong, let's get at the table, roll our sleeves up, and most importantly to me, learn from history. You know, we start screaming about Luddites. Well, Luddites are misrepresented in Western history and the Industrial Revolution to me because they overreacted to save their jobs, but on the other hand, they were trying to protect in their own way things that came about. So child labor laws happened because people organized and said we don't want 12-year-olds losing their fingers. We want to protect workers. So minimum wage, all those things happened because the workforce had a say. Why don't we get ahead of that rather than have people suffer and the industry suffer and have people lose trust as they're losing right now on things like ghost jobs. Thank you, Mr. Chairman. I yield back and look forward to working with you on this. I thank the gentleman for yielding. Now recognize myself for closing statement. I want to thank our witnesses for sharing your perspectives on AI adoption in the workplace. It's been very helpful. AI is changing workplaces across the country and we're going to have to adapt to this new world. As discussed today, we need better data on how AI is changing the workplace so that we can respond based on the evidence. I look forward to working with the members of the subcommittee and the administration to make sure American workers and job creators benefit from the increasing use of AI. Let's make this a win-win decision. And so with that, I thank the ranking member, Mr. DeSaulnier. And before I get further into my closing, without objection, I'd like to enter into the hearing record a statement from Workday on AI and the workforce. So ordered. Okay. Again, thank you witnesses for taking the time to testify before our subcommittee today. And without objection, there being no further business, the subcommittee stands adjourned. [Gavel sounds.]