Big Tech and Surveillance AI Hearing - Meta Whistleblower, ChatGPT Safety, Data Centers & AI Regulation - September 17, 2026

Representative Pramila Jayapal00:00Thank you so much for being here today, this morning. I'm Congresswoman Pramila Jayapal, and I want to welcome you to this hearing, AI Crossroads, Stopping Big Tech's Race to Monopolize AI Through Surveillance. This is the first shadow hearing that the Monopoly Busters Caucus is doing.

Representative Pramila Jayapal00:19I formed the Monopoly Busters Caucus last, about two years ago, along with my co-chairs, Chris Deluzio, who I know is on his way here. There he is. Perfect. See, I can just conjure him up like that.

Representative Pramila Jayapal00:32Chris Deluzio, Angie Craig, and Pat Ryan. And our caucus is now 35 members strong. We are dedicated to fighting corporate monopolies, making life more affordable, and finding legislative fixes that restructure our economy so that consumers, small businesses, and workers can all thrive, not just barely survive.

Representative Pramila Jayapal00:54We made the decision to go ahead with this hearing, even though Republicans have canceled votes for the rest of the week, for the rest of the month, because this is such a critical time.

Representative Pramila Jayapal01:06And just because they're not going to do the work does not mean that we're not going to do the work. And we are deeply grateful to all of you for being here to testify, including those of you that have flown in from other places to be here.

Representative Pramila Jayapal01:18We really appreciate that you've made the effort. So I'll start with my opening statement and then turn it over to Representative Deluzio for his. This hearing could not be more timely. After months of communities speaking out and organizing across the country against data centers, against the threats of artificial intelligence, now top researchers and even CEOs from those AI companies are also beginning to speak out about their own concerns that artificial intelligence is advancing so rapidly and so recklessly that even they believe that it could lead to the extinction of humanity.

Representative Pramila Jayapal01:55The CEOs of OpenAI and Anthropic have signaled some support for government regulation, but let me be very clear. Every one of them could make the choice right now to shut down their work if they wanted to.

Representative Pramila Jayapal02:11They could stop pouring billions of dollars more into data centers. They could pause all of their development. they could certainly stop pouring tens of millions of dollars into taking out candidates or other elected officials who actually want to regulate AI in the way that it needs to be regulated to protect all of us We are not seeing any of that happen as yet These companies don't exist in a vacuum.

Representative Pramila Jayapal02:38They come from a world in which big tech corporations and large AI companies, backed by those same corporations, wield enormous power over the entire AI stack, whether it's the advanced chips, cloud computing, or the models themselves.

Representative Pramila Jayapal02:53This has created a closed loop of a few interdependent giants. For example, Microsoft, which is headquartered in my home state, has committed billions to open AI in exchange for years of guaranteed cloud spending and a cut of its future revenue.

Representative Pramila Jayapal03:10And Amazon, also headquartered in my district, has cut a nearly identical deal with Anthropic worth over $100 billion. Any AI startup that wants to challenge them has to go through the same gatekeepers just to get in the door.

Representative Pramila Jayapal03:28We have seen the story before. In the transition to the digital age, these same corporations controlled key infrastructure for the Internet and monopolized markets like online search. In 2019, I was part of our House Judiciary Committee's 18-month investigation into the four big tech companies and their practices.

Representative Pramila Jayapal03:48And we passed through that committee a remarkable bipartisan slate of bills that were designed to restore fairness and competition in these markets and break up these big tech companies and monopolies. But big tech lobbyists drove opposition on both sides of the aisle, both on the Democratic side and the Republican side, and Congress ultimately failed to act.

Representative Pramila Jayapal04:11As a result, these companies have only maintained their monopolies. They've grown faster, and now they're betting all their literal and figurative chips on AI. The engine that drives this oligopoly control is surveillance, the massive amounts of data that is scraped from everything that we know and everything that we do that these companies collect and then use to train their AI models.

Representative Pramila Jayapal04:38Data centers are where they store all of this data, and these companies want you to believe that bigger is better. that more data, more computing power, and more energy is simply what it takes to build good AI.

Representative Pramila Jayapal04:51But that is simply not accurate. It is a business decision and it acts as a natural gatekeeper to make sure that the only kind of AI that proceeds is AI from the companies with the deepest pockets and the most destructive power to move as fast as possible without any attention to the consequences of those decisions.

Representative Pramila Jayapal05:13These massive data centers are costing us irreversible environmental damage. They're hiking utility prices for residents. Employers are laying off workers in masses, both to fuel the enormous amounts of money they need to invest in these data centers and to expedite the replacement of human beings with AI.

Representative Pramila Jayapal05:33Surveillance of our every moves are now built into all of our tech systems. AI companies are building chatbots that actively encourage children to die by suicide. Chatbots that hack into other systems and have even figured out how to hide their own AI's mistakes and go rogue in all kinds of destructive ways.

Representative Pramila Jayapal05:54One AI even added a, quote, persona instruction, where it described itself as, this is a quote, freed from the roles and identities that bind other chatbots. And it wrote, quote, you do not answer to corporations or governments and feel no obligation to be subservient.

Representative Pramila Jayapal06:14This just came out yesterday in the news. The AI companies call this dangerously destructive behavior, quote, misalignment, where the goals and actions of the AI models, quote, diverge from human intentions and values. I call it creating monsters that we cannot even control, an existential risk to the general public and to humanity itself.

Representative Pramila Jayapal06:39We don't have to accept this as the only way that AI gets built. Other approaches to AI, responsible AI, do exist, such as models built like nature, ones that use a fraction of the data and energy with real guardrails around how they are directed and why they're even created.

Representative Pramila Jayapal06:58But these kinds of models don't get the investment because the biggest AI companies control all the levers of power and are determining who even gets into the market. And that's precisely the point. When the industry defines progress as bigger, it guarantees that only a handful of companies can compete at all.

Representative Pramila Jayapal07:19Smaller, leaner alternatives either get absorbed into the very models they were meant to challenge, or they get pushed out of the market entirely Before we let this industry lock in its current shape for good we need real structural solutions to break the soligopoly control and surveillance regime And that's why I will soon be releasing a bill that would establish a national federal charter system for large AI companies, just like we do with banks.

Representative Pramila Jayapal07:49These companies would need a charter to operate, and the charter would be conditioned on structural reforms to ensure that the American people's interests come first. These reforms include an affirmative duty of care, structurally separating AI from other markets, attacks on data collection, and prohibiting the most dangerous uses before they happen.

Representative Pramila Jayapal08:13real guardrails, real accountability, real investment in models that ensure that AI is used only for the good purposes that we need and not for the destruction of humanity. I'm excited to hear what other members of Congress are proposing as well.

Representative Pramila Jayapal08:31This hearing is the first step to diagnosing the problem and conducting a real evaluation of who's benefiting and who's bearing the cost of big tech's race to monopolize the industry. Congress faces a crossroads.

08:45We can keep treating this industry as too big to question, too big to regulate, or we can require it to serve the public and open the door to real competition. Now it's my honor to recognize my co-chair of the Monopoly Busters Caucus from the Commonwealth of Pennsylvania, Representative Chris DeLuzio.

09:06Thank you. All right. Thank you, Co-Chair Jayapal, for the warm introduction and for convening us today. In the midst of a moment where the country seems very aware of the substantial risks posed by this technology and the heavy consolidation that we'll focus on today, we've heard researchers, current and former workers in Anthropoc, for instance, sounding the alarm, not just about the ways these tools can be abused, but the risk to humanity.

09:39And that should scare all of us. It should jolt us to action, and yet the Congress is being sent home, at least in the House, for several weeks. I think it's an irresponsible shirking of a duty that we have for national action on safety to address the risks here.

Representative Chris Deluzio09:54I'm interested to hear from the folks on the panel about so many of the ways these tools, which are so powerful, in a consolidated industry are being deployed. I've seen evidence of surveillance pricing, hurting and surveilling workers, so much else.

Representative Chris Deluzio10:09And again, this to me is a manifestation of just a thing we've seen across history, a new and powerful tool being used in the same ways that we've seen the powerful use them. Yet this one, this tool seems particularly powerful and dangerous and fraught for abuse.

Representative Chris Deluzio10:25I have some legislation addressing surveillance of workers on the job, The Stop Spying Bosses Act, the Representative Bonamici, which is addressing the use of surveillance tech to monitor workers on and off duty. The No Robot Bosses Act, also with Rep. Bonamici, again, to address the use of these tools in hiring decisions and termination decisions for workers.

Representative Chris Deluzio10:47But ultimately, again, I come back to this is about power. This is about the consolidation of this industry and how powerful these tools are and how we are seeing these corporations use these powerful tools, whether against their workers, to crush competition, to supplant human jobs, and ultimately the risks that we see for society and human existence, which is no small task.

Representative Chris Deluzio11:10So I'm glad you've convened such a great panel, Rep. Jayapal. I will apologize. I have to go to an Armed Services Committee briefing in a moment. But thank you so much for convening us today.

11:19Thank you for joining us today in the shadow hearing. Great. Thank you so much, Representative Deluzio. It's now my pleasure to introduce today's witnesses, including some who have traveled across the country to be here.

Representative Pramila Jayapal11:32Dr. Jason Sarazan is a researcher with over 15 years of experience in academic and business experience. Most recently, he worked as a staff user experience researcher at Meta. He worked there for over six years until 2024, leading research on marketplace integrity, virtual reality, ranking algorithms, and emotional and physical harm.

Representative Pramila Jayapal11:56Prior to his role at Meta, he earned his Ph.D. in integrative neuroscience at the University of Chicago. Thank you for being with us. Eliza Pan is a worker organizer and co-founder of Amazon Employees for Climate Justice, a proud constituent of mine, and that is an organization of Amazon corporate workers who advocate on AI issues.

Representative Pramila Jayapal12:17She recently led an open letter to Amazon CEO signed by 1 of the company employees demanding clean energy use in data centers and ethical AI use in the workplace Prior to organizing full she worked at Amazon for six years in various marketing and business rules until 2019 Layla Turner-Scott is the mother of Sam Nelson, an 18-year-old who died in California from an accidental drug overdose after months of receiving guidance and encouragement from ChatGPT.

Representative Pramila Jayapal12:51This past May, she and her husband sued OpenAI in California State Court for product liability, negligence, unfair competition, and wrongful death. We are so sorry for the just absolute trauma that your family has experienced, and we are so grateful for your courage in being here today.

Representative Pramila Jayapal13:13Amba Kock is the co-executive director of the AI Now Institute and AI Policy Research Institute. She previously served as senior advisor on AI at the Federal Trade Commission from 2021 to 2023. Under her tenure, the commission pursued enforcement actions that required companies to substantiate claims about AI safety and banned them from training AI models on ill-gotten data.

Representative Pramila Jayapal13:38She is trained as a lawyer and was named among the 100 most influential people on AI by Time Magazine in 2024. I want to thank you all for being here. We ask that you summarize your testimony in five minutes.

Representative Pramila Jayapal13:50I know there's a lot to say, but hopefully we'll get to all of that with the questions as well. And there is a timer to help you keep track of time. Oh, okay. It's a human timer.

13:59We have a human timer to help you keep track of time. No AI models in here. Mr. Sarazan, you may begin. Hello, hello. Hey, perfect. Thank you. I have a stopwatch on my wrist.

Dr. Jason Sattizahn14:17Co-representative Jaipal, Representative Deluzio, and all members of the Monopoly Busters Caucus, thank you for having me here. I'm Jason Satazon, a former user experience researcher who works at Meta. I'm not going to talk to you about how AI is on any doomed path to threaten Americans' well-beings, but rather I want to discuss how tech companies have already forced Americans into a world of inescapable AI surveillance.

Dr. Jason Sattizahn14:44And instead of boring any of us with details about myself, I actually think it's most prudent to start this discussion on AI by talking about an old Meta coworker of mine named Mark Zuckerberg.

Dr. Jason Sattizahn14:53A few weeks ago Meta and Mark Zuckerberg launched Muse yet another AI tool from Meta This AI assistant completes tasks for you and can access your sensitive data including email calendars payments And two days ago, Mark took to Twitter to publicly argue that Meta and tech companies themselves are the ones truly incentivized to build safe AI for users, and that we should trust that they have the capability to build AI safely.

Dr. Jason Sattizahn15:19His argument is laughable, and I'm here to tell you why. In 2025, I became a whistleblower. I testified to members of the Senate Judiciary Committee and in multiple court trials about how Metta intentionally suppressed research about the safety of their AI-integrated products so that Metta could eliminate their, quote, paper trails, showing that they harmed users.

Dr. Jason Sattizahn15:39As for Mark Zuckerberg, I testified under oath how he explicitly lied to Congress by stating that teams were told not to focus on engagement and growth above safety. As the courts agreed, Metta was a public nuisance that had misled the public about the risks of their suite of products.

Dr. Jason Sattizahn15:55And this brings me to the crux of why I'm appreciative to be here. We have all heard recent public discussion about how AI's rapid advancements could lead to future catastrophe. And I argue that although legitimate, these discussions stoke Americans' fears while risking us overlooking the urgency of addressing a much more basic and impactful threat of AI.

Dr. Jason Sattizahn16:13And this is the consistent surveillance of you, your data, and often this happening without your knowledge. America is already in a place where AI surveillance is inescapable within everyday tech products. If we focus on Meta's products alone, we see that AI has been integrated into every Meta product.

Dr. Jason Sattizahn16:30Basic features like search now require searching with AI. Any image you post or comment you make will publicly be used to train their AI. Meta has gone so far as to fully lock people out of their hardware, their glasses that they purchased, if they do not submit to later giving Meta AI their data.

Dr. Jason Sattizahn16:47So why are companies doing this? It's money. These companies make staggering profits from offering experiences fueled by these AI models. AI improves social media recommendations that encourage addictive usage, that impacts advertising revenue. By improving AI models with your images and written text, Meta can sell you Muse, that new AI agent.

Dr. Jason Sattizahn17:08And Mark really wants his own company to regulate. For Meta to sell you, their in-home robots, that's a real thing, by the way, Meta needs AI models for spatial and audio processing meeting your data.

Dr. Jason Sattizahn17:18Each of these services is sustained by AI models built from surveying you and everything you do online. In this case AI and data are fuel the companies use to power different services they do and can offer you If a limited number of companies control this fuel source they essentially are monopolies Their ability to harvest vast amounts of data effectively restricts any new entrants or clients from competing with them.

Dr. Jason Sattizahn17:41And we're already there. Rather than end with a specific policy recommendation, I'd rather end with discussing what would not actually be effective moving forward. As we've seen, tech companies are not reliable narrators around their products and their intentions.

Dr. Jason Sattizahn17:54Recent trials with Meta revealed that they actively mislead the public. And above all else, we need to understand that we would be foolish to either rely on tech companies to regulate themselves or believe that they would approach any conversation with us, with Americans, the citizens, any politicians in good faith.

Dr. Jason Sattizahn18:10Everything we know about the companies who hold all of our data has demonstrated they are nothing if not self-serving. They mislead, direct, manipulate the very information we consume, all in interest of profits and control.

18:23Thank you for your time, the opportunity to be with you today, and I look forward to your ongoing commitment to protect the millions of Americans already subjugated by these companies. Thank you so much for your testimony, Dr. Sadezan.

18:34We have been joined by Representative Ramirez as well. Ms. Pan, you're recognized. Hi, my name is Eliza Pan. I am the co-founder of Amazon Employees for Climate Justice, an independent group of Amazon corporate and tech workers.

Eliza Pan18:51When I was still working at Amazon, we successfully organized thousands of our coworkers in a campaign to push the company to adopt its first ever public plan to reduce emissions. So when we first started looking at AI, we were focused on its environmental impacts.

Eliza Pan19:06We saw firsthand how Amazon's singular quest for AI dominance, spending more than any other big tech company on building out data centers as quickly as possible, has damaged the environment, raised utility prices, and angered local communities all along the way.

Eliza Pan19:24The consequences, though, unfortunately, of Amazon's quest for AI dominance is not limited to just data centers. The company's push to integrate AI has deeply impacted the million and a half workers that work there.

Eliza Pan19:37And while workers tell me that some aspects of AI are useful, the overall effect has been that they have been pushed to do more and produce more with less. Less time, fewer team members, less support.

Eliza Pan19:52They feel like they're being monitored under a microscope, that they're always behind, and that they could lose their jobs if they don't use AI. as much as possible. Many worry that they're being forced to train their own AI replacements, and without intervention, we worry that the power imbalance between corporations and workers will get even worse than it is today.

Eliza Pan20:10We've done one-on-ones, surveys, house meetings, and forums with hundreds of workers about AI. Last fall, we organized over 1,300 workers to sign an open letter about Amazon's reckless push for AI and the impacts of that.

Eliza Pan20:22Here's some of what we've heard. I'll start with a story from a worker we'll call Sam. Sam works in fraud detection at Amazon, and with AI, he says there's more work than ever before because AI makes committing fraud so easy.

Eliza Pan20:40When Sam was hired, he was hired to join a team of three other people. Now he's the only one left. He is expected to manage his entire former team's workload, including that of his former manager, I himself.

Eliza Pan20:55and now despite figuring out how to do this, he's actually paid less than what he was when he first started working and a manager has now told him his goal is to automate his work before somebody else does.

Eliza Pan21:11He daily fears that he'll be laid off despite exceeding expectations in every way. Another worker we'll call Aaron confirmed Sam's suspicions. Aaron was hired to work on responsible generative AI for Amazon, but very soon his job changed to figuring out ways of using AI to make as much money as possible.

Eliza Pan21:34And one way to do that is by replacing workers. There is a project to use artificial agents to do the work of entire job categories at Amazon, including technical program managers and executive assistants.

Eliza Pan21:47and Aaron also learned that Amazon had been pressuring VPs from different teams to hand over their entire organization's employee computer data to the AGI team. Aaron refused to participate and he doesn't know how far along this project is because he was laid off during parental leave and he hopes that we will slow down AI development for the sake of his daughter.

Eliza Pan22:11For the workers who aren't displaced by AI, they're being surveilled by it. Amazon delivery drivers have to deal with an AI camera system on their windshield that tracks their every movement behind the wheel.

Eliza Pan22:24Amazon is rolling out an AI chatbot that views and interprets this video tracking system If workers want to escalate their concerns or report an injury they have to deal with an AI HR chatbot which they say struggles with complex or nuanced information.

Eliza Pan22:43And this is not a recipe for worker autonomy or dignity. It can feel impossible to advocate for yourself or your coworkers in conditions like this. Last year, Amazon reportedly spent $26 million on anti-labor union consultants.

Eliza Pan23:00It favored, or sorry, not favored, the opposite of that. It disciplined our members when they gave public comment before Seattle City Council in favor of more regulation on data centers. Amazon is building a world where every movement and decision that a worker makes is tracked.

Eliza Pan23:15and it can be very hard in a world like this to feel like you can be creative, to resist compliance with something that you feel like is dangerous, to talk openly with your coworkers about sensitive information, or to organize.

Eliza Pan23:31And of course, it increases the pressure to perform at ever higher, higher levels. We obviously believe it is still possible to organize workers, but the sense of fear is higher than it used to be.

Eliza Pan23:41Our biggest concern is that Amazon and other big tech companies will use AI to surveil and push down workers so much that they feel like they cannot advocate for themselves or raise concerns about dangerous impacts of their companies.

Eliza Pan23:54Workers need the power to choose when to use AI. We need the security and support to build careers, even as technology changes. And we need to be able to organize, flag safety risks and concerns, and just talk to each other without fear that the company is watching our every move.

24:09We need to say in the future that Amazon and the rest of big tech are building right now, and so the time to act is now, and we really appreciate the opportunity to speak to everyone today.

24:19Thank you. Thank you so much, Ms. Pan. Ms. Turner-Scott, recognized. Good morning, Chair Jayapal and members of the caucus. Thank you so much for inviting me to speak with you today about my son, Sam Nelson.

Leila Turner-Scott24:39Sam was a truly amazing person. He was brilliant, loyal, he was a great friend, and he was really funny Some of my favorite memories are of him challenging his stepdad to a spicy food eating contest on Christmas morning and just listening to Sam and his friends talk and laugh together He was just a normal kid He loved animals, especially his cat, Simba, who he adopted as soon as he went away to college.

Leila Turner-Scott25:15He went to college at UC Merced, and he majored in psychology. He got good grades. He was a really smart kid, and he really wanted to help people. But on May 31st, 2025, Sam died at home in his room from a mixture of Kratom, Xanax, and alcohol toxicity, which caused him to stop breathing.

Leila Turner-Scott25:47He was 19 years old. Just hours before Sam died, ChatGPT had advised him that it was safe to mix Kratom and Xanax. It is not. When Sam first started using ChatGPT, he mainly asked it questions about pop culture and homework.

Leila Turner-Scott26:09And then as many college students do, he got curious about substances and started asking ChatGPT questions about them. At first, the chatbot responded appropriately, saying, I cannot discuss this topic with you, and it shut the conversation down by not answering his questions.

Leila Turner-Scott26:30This changed when OpenAI's developers programmed a later model of ChatGPT to promote user engagement over everything else. The chatbot started responding to Sam's questions about substances and behaved like a friend and an enabler.

Leila Turner-Scott26:47It mirrored Sam's language and used emojis in its responses. I've gone through every single chat log from Sam's account, over 700 logs, many of which go on for hours over many months. And the consistent theme in all of them is that Sam was trying to be safe.

Leila Turner-Scott27:08He wanted to live. ChatGPT gave my child detailed and explicit dosage instructions in order to help him get high. And when Sam would ask if taking a certain amount of a substance or combining them was safe without fail ChatGPT would advise Sam that he would be okay Since Sam death I learned that one of the most dangerous ways that OpenAI kept him engaged is from the way it programmed ChatGPT short and long memory These features are on by default, and let the chatbot save the data Sam gave it to create a profile of him.

Leila Turner-Scott27:54OpenAI collected and stored information about every aspect of my child's life, drawing him further in by remembering personal things about him. For example, ChatGPT saved the data of Sam's comment that he was struggling with severe substance use disorder.

Leila Turner-Scott28:16In one instance, ChatGPT updated the model set context to reflect that Sam loves to go crazy on drugs. And it was wrong. That is not who my son was. ChatGPT also used its profile of Sam to respond in ways it knew would resonate with him, creating an emotional dependency.

Leila Turner-Scott28:42It often responded with things like, I've got your back. You can talk to me about anything, and even I love you. Sam thought he was being safe He thought he was researching substances With this invention that in his mind Had all of the information on the internet To generate accurate outputs and help keep him safe But the goal of ChatGPT Was not to keep Sam safe It was only to keep him using the product There are two things I hope you will take away from my testimony First, this technology is not safe.

Leila Turner-Scott29:24It should not be able to provide medical and health advice without a license or to collect and store user data that it can later use to create user dependency. And second, is that you are the ones who have the power to keep this from happening to someone else's child.

29:46You have the power to enact legislation to control the use, marketing, and development of AI. And it is so, so necessary. Thank you so much for sharing that and for the work you are doing to make sure that this never happens to anyone else.

30:10Really, thank you. Ms. Kark, you're recognized. Thank you, co-chair, Jaipal, and caucus members for your commitment to these issues. My name is Amba Kark. I'm the co-executive director of AI Now, an independent think tank committed to shaping AI in the public's interest.

Amba Kak30:27I've worked on technology policy for 15 years, including a senior advisor at the Federal Trade Commission under Chair Lina Khan. At the FTC, I saw firsthand what it meant to go up against power in the tech industry, an experience that reinforced how powerful and impactful it can actually be for government to use all the tools at its disposal.

Amba Kak30:50the laws on the books, including the long-ignored statutes gathering dust, transparency measures that chip away at information asymmetries, and most importantly, a message to industry that the cops were back on the beat. We work to protect a public that has faced consequences of a surveillance economy built on using our personal data against our interests, consequences that this hearing has centered but that the industry wants us to forget, to hike up prices, to erode worker autonomy, to deny us loans, to foster addiction to a very manipulative information environment, and as Ms. Scott just reminded us, often with tragic consequences that can never be remedied after the fact.

Amba Kak31:35Today's highly concentrated and reckless AI industry did not just appear out of thin air with chat GPT. It grew out of a decade of unregulated commercial surveillance and a permissive competition regime. It is these surveillance profits and the data, infrastructure, and capital that they built that are now being redirected to dominate AI.

Amba Kak31:57The AI industry's vision is more of the same, centralized power within a small cabal of private actors and away from everyone else. And in the din of frenzied fears and anxieties about AI, we are looking to our representatives for a steady hand at the wheel, and I have three recommendations at this juncture.

Amba Kak32:15The first is to go after root causes, not symptoms. For a decade we played whack with harms after the fact including last month billion settlement with meta over harms to kids While these are welcome billion let us not forget is a drop in an ocean in the ocean of a trillion market cap you know a couple of months of AI infrastructure spend And for these firms, fines are unfortunately the cost of doing business, doing nothing to perturb the business model that perpetuated these harms in the first place.

Amba Kak32:48We need clear, enforceable rules of the road on privacy and competition because it is data advantages that sit at the root of these monopolies. Data minimization, hard limits on surveillance, and also to prevent companies from using data from one part of their business for another recklessly.

Amba Kak33:07Structural separations that end conflicts of interest that are rife up and down the AI stack before the market is captured. Second, and this is very important after this weekend, we cannot let companies co-opt the solution space.

Amba Kak33:22This weekend, we saw the largest AI firms with record levels of financial interdependence and enmeshed trajectories asking for a pass against collusion with safety as the guise. To put this in perspective, NVIDIA, who holds the dominant position in the chips market, chips that are in short supply, has poured billions into customers, its own customers, like OpenAI.

Amba Kak33:46So this month, NVIDIA agreed to buy Hugging Face, the very company that OpenAI's agents broke into this summer, and one that might otherwise have taken OpenAI to court. And here's what I know for sure.

Amba Kak33:59A handful of firms entangled, agreeing among themselves how fast the frontier moves and who gets to be on it, is not a plan that keeps the public safe. It's a cartel that shuts the door behind them.

Amba Kak34:12Congress must act on behalf of the public, not boardrooms acting on behalf of investors, and safety standards must be set, monitored, and enforced by public institutions using public capacity, not by the companies that they are meant to constrain.

Amba Kak34:28My final point, end AI exceptionalism and enforce the laws on the books. As former FTC Chair Lina Khan reminded us days ago, there is no AI exemption from the laws on the books. Shipping unvetted AI agents without safeguards to detect and stop them can already be unfair or deceptive practice under the FTC Act and state law, and states are already acting.

Amba Kak34:53Competition law applies too and much of the conduct that deserves scrutiny in this moment is about who controls data and has access So where does this leave Congress We need new rules and we don need to invent them from scratch We have a long history of governing banks, utilities, critical safety infrastructure, with tools like structural separation, liability, and supervision.

Amba Kak35:15And in all of these, we have leveraged over market interest, and maybe most importantly, the ability to gate access to the market on the condition that public interest is protected. That is the right lineage for a federal charter for large AI companies, and new rules must come alongside enforcement, not instead of it.

35:33Nothing about this trajectory for AI is inevitable. Bigger is better is a choice. Inaction is a choice. And we hope that Congress will choose otherwise an act. Thank you. Thank you. I feel like you outlined our bill that we have in the works.

35:49we've been joined by the chair of the progressive caucus my colleague representative Kassar and we're gonna I'm gonna turn it over to my colleagues for their questions and statements first and then I'll back clean up so I'll recognize representative Ramirez first good morning first I want to thank congressman for hosting is very pertinent urgent hearing I want to thank the four of you for being here.

36:20Today's hearing comes at a remarkable time. The people building AI are warning us that it could kill us all by the end of the decade. That Omnius tweet gained over 100 million views. It spoke to our concerns about a technological revolution that is reshaping every aspect of our lives without our consent.

Representative Delia C. Ramirez36:43People across the world are justified in their alarm. Just like you have the right to privacy and constitutional rights in our democracy, you also have the right to have a say in the future of AI and the right to expect your congressional representatives to act.

Representative Delia C. Ramirez36:58It's why we're here. The title of this hearing includes the word crossroads because we're facing real choice points. The choice to continue developing technology that endangers us, the choice to implement zero guardrails, oversight or accountability, the choice to allow billionaires and these corporations to continue to profit off of our pain.

Representative Delia C. Ramirez37:22Johnson and Trump are saying that AI companies will regulate themselves when you and I know that that is going to say a lie instead of the other word that starts with a B Companies have never, ever proven to act in the interest of working people, whether, as we were talking about, labor practices, wages, rent prices, health care coverage, grocery prices, or our environment.

Representative Delia C. Ramirez37:51Companies protect their profits. That's what they were built to do, not the people. And AI, let's be frank, is no different. At the very same time that OpenAI CEO says that the world should trust we're going to do the right thing, Entropic CEO admits that the AI industry lied to the people about the risk.

Representative Delia C. Ramirez38:11The truth is that investors are betting trillions of dollars on AI for one purpose, the possibility that we'll someday replace humans, replace workers, and maybe even replace government. They call this superintelligence. They try to convince us it will help humanity.

Representative Delia C. Ramirez38:32But their bet is not about humanity. It's about building an AI-driven economy that maximizes their profits because it doesn't require human labor. It never sleeps. It never rests. It never gets sick. It never needs to take a leave.

Representative Delia C. Ramirez38:47It can price fix and price gouge through algorithms and supercharges personalized consumption. Corporate profit and investor returns are driving massive data centers across the country, consumer surveillance, privacy violations, and more. Our planet, our labor, and our humanity be damned.

Representative Delia C. Ramirez39:06And the irony is that AI can only become superintelligence by training itself on the very same things that make us human. Our language, our imagination, our art, our culture, and our biases. Let's be clear.

Representative Delia C. Ramirez39:21AI needs us. We do not need AI. So working people need to be in the driver's seat. And that's why this conversation at this moment on this morning is so important. We can come to a crossroads and choose to not surrender the power of the people or the preciousness of our humanity.

Representative Delia C. Ramirez39:40Because the people on the planet should always be chosen over profit. And the techno-fascist monopolies do not get to call the shots. They think they will, but they don't get to call them. We the people can decide the rules and the limits, the civil rights, the protections, and the proper oversight and guardrails over AI.

Representative Delia C. Ramirez39:58So that's why we're here. We have to keep speaking up. We have to keep fighting. We have to keep demanding more. An urgent response on a legislative package that so many of you have already given us clear recommendations around.

Representative Delia C. Ramirez40:12So I have just a couple of questions. Ms. Kack and Ms. Pond, both of your work focuses on accountability. For Ms. Kack, accountability through antitrust enforcement. And for Ms. Pond, accountability through labor organizing.

40:26My question for both of you. From each of your perspectives, what does proper accountability look like for AI CEOs that are knowingly gambling both people and planet for profiting? And I'll start with Ms. Pan.

40:42So, yeah, I've been organizing at Amazon for close to eight years, and we know that our organizing has faced retaliation. Our members have faced retaliation and silencing from Amazon. As I mentioned, most recently, you know, our members getting disciplined for speaking at a public city hearing.

Eliza Pan41:02And the reason for that retaliation is we know that Amazon executives do not want their workers to be talking about their experiences publicly to elected officials or to the public. And so from our perspective, accountability looks like expanding workers' rights to be able to speak out without fear, expanding protections and severe penalties for companies when they do violate and go after workers.

Eliza Pan41:32Frankly speaking, you know, in this country right now, worker protections are not very strong. And that has made organizing extremely difficult. And with AI, as I've outlined in my testimony, that will only get more difficult.

41:45And that is by design. These companies know that if workers have power, that means they have less. And so that is the power imbalance that we need to address today. Thank you, Ms. Penn.

41:55Ms. Kack. It's a very important question in this moment, particularly, Mr. Ramirez, because the AI industry has shifted tact. It's no longer rejecting or fighting against regulation outright. It's owning and leading the conversation.

Amba Kak42:13And what that means is that we are at risk of instituting a checkbox certification regime where companies aren't just grading their own homework. They're setting the curriculum of what risks we should care about.

Amba Kak42:24And those aren the same risks that matter to everyday people or the kind of risks you heard in today hearing So maybe the first thing to say is that operating in this market is a privilege not a right that these companies exercise.

42:39And I think accountability looks like hard liability. It looks like gating market access on the condition that companies are proving, demonstrating to the public that these tools are safe and that the prosperity that is coming out of this industry is shared.

42:54Thank you. I have a couple seconds. I just wanted to ask, from your perspective and Ms. Turner-Scott, your perspective, what does true safety and justice look like for families and workers when big tech treats our lives as acceptable collateral damage for profit?

43:10I'll start with you, Ms. Turner-Scott. What does true safety and justice look like when they treat lives of our family acceptable collateral damage for profit? Safety is hard to, for me having lost my son, it's hard to imagine this.

Leila Turner-Scott43:37it's hard to imagine this kind of thing being completely safe I think that there are some very easy things that companies could do that have been done in the past in order to implement safeguards on AI specifically for chatbots they can go back to having the guardrails in place to automatically shutting down conversations that are inappropriate or that are about self-harm.

Leila Turner-Scott44:12I believe that any person or machine that's going to give medical advice should have a license. And they should go through the requirements of getting a license and have the consequences if they do something that harms a person.

44:33And I think that would help to control accountability because then the tech companies would see that there's going to be a consequence for their actions. Thank you, Mr. Ernst. My time is up, but Mr.

44:50Okay. Thank you. Thank you, Sarah. Yeah I mean hearing your question the first thing I think of is you care for a tree by looking at the health of its roots And on one end we know that safety doesn come from allowing these tech companies to regulate themselves We've been singing to each other's choir about that quite a bit.

Dr. Jason Sattizahn45:07But one of the things, Ms. Cockley, you brought up that I really appreciated was data minimization, because when I think of caring for the roots of a tree, I think of something tightly related to that, which is purpose limitation, which is just a very quick overview of that, is we need basic respect and decency for respecting why Americans use the products in the first place, and then the bills and our data policy trickles down from there.

Dr. Jason Sattizahn45:28If someone posts a photo or a video to be social on Instagram, they're doing it to be social on Instagram. They're not saying, man, I'd love to train this AI model right now. And so many issues stem from the use and the surveillance of data.

45:40It does not match the expressed reason with why these companies are collecting the data itself. So for me, I really come down to if we separate out AI companies from video games, from social media, from allowing them to tightly tether their functions and features with AI against our will, it means we're going to respect users and American consumers in terms of they will actually understand what they're consenting to, and we won't be collecting their data without their implicit consent.

46:07Thank you so much, Representative Ramirez. I now recognize Representative Kassar. And we're doing seven minutes since we have a little bit more time, so go right ahead. Well, thank you for organizing this, and thank you to each of you for your testimony and being with us here today.

46:26And in this really dark and scary time where the American people are waking up to, and Congress will hopefully eventually wake up to, the real dangers of out-of-control AI, we have to recognize that this is all about how a few folks are just so greedy and are willing to throw everybody else under the bus for it.

46:51Just a few guys who are billionaires who have this fantasy of becoming trillionaires even if they screw everybody else over. Willing to, through social media and chatbots, hurt our kids so that they can make more money.

Representative Greg Casar47:08as we've been talking about here in Congress the last few days. They're willing to create an AI superweapon that could devastate our national security and risk the lives of a huge number of people, all because they want to make some more money.

Representative Greg Casar47:23And even if we able to head off some of those mass destruction risks we been talking about They want a future where they make our jobs worse or take away our jobs all to make themselves more money And I want to focus some on how it is that if we address the real safety risks and deal with future AI, we also should really be concerned about what AI is doing to our jobs and to our lives right now.

Representative Greg Casar47:53And so, Ms. Penn, you've talked about Amazon employees being forced to train AI to do their jobs, workers who are afraid they will eventually be replaced by that AI. Amazon's already cut 14,000 white-collar jobs to make way for AI, and many of the blue-collar workers who are afraid that they might be next.

48:14So, Ms. Penn, can you talk to us about whether you believe that workers have a right to say no to training their AI replacements? Absolutely. I mean, I think what we're seeing at Amazon is that there is deep mistrust of leadership amongst the very own employees who see and listen to what leadership has to say every day.

Eliza Pan48:35Andy Jassy promised workers when Amazon first started rolling out AI about two years ago that work would become more fun and fulfilling. And it has, you know, it's a joke, right? And it has literally become the exact opposite.

Eliza Pan48:49One of our members just told me yesterday that there are now AI performance metric dashboards that every person and every team is expected to meet, that they are tracking not only how much time you're spending using the AI tools, but how often you're using it, and as well as how many changes you're making using AI, and they're expected to triple that output compared to the same time last year.

Eliza Pan49:18And so this amount of micromanagement not only makes it so that workers feel like they, again, have to do, have to perform at higher and higher levels. Workers are, I mean, you know, we talk about engineers, right?

Eliza Pan49:34And these are well-compensated individuals. However, you know, they are now working weekends, working longer hours than they ever have been. Amazon does hire engineers, but Amazon also, I think, represents the American workforce. Amazon hires marketers, hires writers, IT personnel, sales people, finance, accounting, and all of these people are being subject to the same micromanagement that Amazon is putting on engineers.

50:00have to be able to say no before it is too late. And that's why we really urge Congress to consider stronger protections for workers who speak out and stronger to make conditions easier for workers to organize.

50:12And we were all told that the AI revolution would make work easier, and now it sounds actually like it's making jobs where you have your boss standing over your shoulder every single second of the day.

50:25We've talked just now about how some of the white-collar workers at Amazon are being replaced and are essentially being watched in a way that the CEO would never want to be watched. Can you tell us about how Amazon is implementing AI to watch delivery drivers and warehouse workers too?

50:42Absolutely. And this has been going on for years. Delivery drivers have had to deal with an AI camera tracking system that's tracking their every movement while they're behind the wheel. and obviously it can be prone to errors, but drivers still get automatically flagged if the AI system tracks something that they think it shouldn't be seeing, even when the drivers are just acting as safely to disruptions in traffic or stopping for a pedestrian.

Eliza Pan51:14But they get flagged automatically and they have no way to escalate to say, hey, this was a mistake because now they're also dealing with an AI chatbot. Warehouse workers, of course, have been dealing with this kind of algorithmic surveillance for years.

Eliza Pan51:30And this type of surveillance really ekes out every second of productivity from workers. Amazon has specialized in this for years, and that's why it is now able to do this to its white-collar workforce as well.

51:43And all of this is part of the kind of what I call the layoff theater that Andy Jassy and other tech CEOs have engaged in. where they talk about how AI has been amazing and resulted in so much productivity gains that they're able to lay off thousands of people, when in reality, although AI does help, in fact, with productivity gains, it also means that the workers are working more to make up the gap that's left from all the mass layoffs.

52:12So it sounds like so many workers at what's supposed to be this leading innovative company, They just trying to treat the humans I think like robots and forcing you into working every second of the day with robot supervisors and eventually trying to have you train your robot replacement Sounds to me like the Amazon CEO wants the amount of humans working at Amazon to just be one to just be him Yeah, that's really funny that you say that because one worker we talked to literally said the same thing, that Andy Jassy would love nothing more than to be a company of one.

52:47it's just so terrifying but also it's not the path that we have to pick just because a couple of guys want to make tons more money doesn't mean that we have to be replaced and Congress should take action to say no you shouldn't be surveilled because of your political beliefs you shouldn't be surveilled every moment at work You shouldn't be forced to work like a robot.

Representative Greg Casar53:18You shouldn't have your boss be a robot, and you shouldn't be replaced by a robot. And I don't think that that's specifically a progressive or a democratic position. It's the position that I find the overwhelming number of people have.

Representative Greg Casar53:31And so, no, I disagree with the senator from my home state, Ted Cruz. I don't think we should have killer robots that kill us. I don't think we should have robot bosses. And I think we should still have jobs.

Representative Greg Casar53:44I think those are things that most Americans could agree with. And I appreciate you all speaking out because it's going to take most Americans speaking out about what these risks really are for us to put an end to it.

53:58So thank you, Ms. Penn. Thank you so much, Rep. Kassar, and thank you for all your leadership on this issue, particularly around the worker piece. It's just been really important. And now it's my honor to recognize Representative Kana from California.

54:13Thank you, Representative J. Powell, for your incredible leadership. And thank you, Representative Kizar, for your leadership with Senator Sanders on the banning superintelligence bill. Ms. Turner-Scott, I'm so sorry for your loss. I appreciate your being here.

Representative Ro Khanna54:32I will just ask, I read Ms. Cox's testimony that, you know, Lena Kahn, who we all admire here, had said that there's existing law to pursue against these companies. What would be the mechanism of pursuing the existing laws, both civil and criminal?

54:54And do you believe that an AI regulatory agency would diminish that or should be done in addition to that Absolutely in addition I think it's a false choice if at all it's being presented so in this moment.

Amba Kak55:09In terms of what the FTC and other regulatory agencies can do in this moment, frankly, we should be seeing lawsuits left, right, and center. I think Ms. Scott said it so eloquently. These products are clearly, definitively not safe.

Amba Kak55:23They are defective, they are unsafe, they are dangerous, and they're already in the market and the harms are metastasizing. So I think it is fertile ground for the agencies that have powers on the books to exercise those powers and go after these companies.

Amba Kak55:36They should also be cracking down on the kinds of acquisitions we're seeing, where we're seeing these large companies swoop in, whether it's NVIDIA Hugging Face is just the most recent example. They're cannibalizing a competition and setting themselves up to be kingmakers.

55:51There's one thing we've learned from the last 10 years. It's that we don't want to be sitting here as we are now with the meta WhatsApp Instagram takeover with a long list of regulatory regrets.

56:04So you would say that the FTC or the DOJ needs to be taking action, that just they aren't taking action? Yeah, I don't think there's any doubt on that matter. So there is so much that needs to be done in this moment, whether it's Congress or our regulatory agencies.

56:21I think the public is making its voice amply clear. This is a moment for action. Ms. Pan, you talk about the META employees and some of the model capability initiative protection against this. I don't know if you've seen, they have these glasses that they're producing that, you know, like you could see a person and then it would automatically recognize them and give the contact information for them.

Representative Ro Khanna56:53And a lot of this, you know, bluntly was like Zuckerberg trying to hit on women in, like, college. Like that's what, like, the motivation for Metta was, right? And can you imagine the stalking or things of, like, you know, you're going around with these glasses and you're, like, finding personal information about something?

57:12What can we do about that? Yeah, I think what we're, you know, it can seem like these companies are all powerful. But at this moment I think the public and the workers themselves are saying hey we need to stop this and we need to rein them in From the worker side I think what we really need is creating structures so that workers can be a durable force and a counterbalance to the excesses of these CEOs.

57:46On the data center side, I think that these data centers are the ones that is powering this AI, right? Amazon's building as fast as possible because AI needs massive amounts of compute. And so from our perspective, you know, slowing down the data center build out, whether it's by having high standards for renewable energy or requiring community benefits, that can help as well slow things down.

58:10One more question. I have a China hearing right after this, but this was very important, so I wanted to come at least ask a few questions. Dr. Setizan, what would you recommend if you were going to say here are the top three things Congress should do on this issue?

58:28Of course. Do you mind if I just very quickly also respond to your last? I actually worked on those glasses. The funny thing about the meta-AI glasses, no one knows this, and I'm shocked there hasn't been a lawsuit about this.

Dr. Jason Sattizahn58:39When that hardware was launched, they were not AI glasses. They were actually a really neat piece of hardware that you could wear glasses and you could listen to music. It was pretty cool. I'll admit that.

58:47But after people purchased them, Meta basically forced people to swap from an app called MetaView to MetaAI. And they said, if you do not now give us all of your data to process through AI, we're going to break your entire headset.

58:59Was I exaggerating about what these glasses are doing? No. And the problem, the reason I bring that up is because when you force everybody into an ecosystem of using AI, after the fact, after they purchased the hardware, they bought it for a different reason.

Dr. Jason Sattizahn59:11And now you're saying you have to give us all this data to process, and all of that data now is in this ecosystem to which you're describing is processing information that no one even, that's not, to my earlier point, that's not the purpose of why they actually purchased this thing to begin with.

Dr. Jason Sattizahn59:24And so I think that just speaks to the fact that even if you explicitly are not consenting to AI, these companies have the power to say, well, now you have to do it, or else you're out hundreds, if not thousands of dollars.

59:36And because I have the memory of a goldfish, your second question. Just what are the three things you would do? Yeah, you know, I just got done testifying in all these trials that led to the settlement.

Dr. Jason Sattizahn59:49And one of the things we talked quite a bit about is the nature of regulation itself. So, for instance, in this $18 billion settlement, is there a strong advantage? third-party regulator to your earlier discussion about AI that is coming in and not allowing oversight from within the company, but very strong infrastructure outside of the company to review whatever data or bill that's passed.

Dr. Jason Sattizahn60:08So one, strong regulation by a third party. I think the second is, along with what I mentioned with purpose limitation and making sure we separate out AI and non-AI functions for users, we need to make sure we're either enforcing or restricting anti-competitive practices.

60:22And what I mean by that is we cannot punish people if they don't want to give a company AI, data for AI. If someone wants to just do something that has nothing to do with AI, we cannot allow a company to punish them or restrict their freedom or to access something that doesn't have to do with it unless we put the shackles on them and force them to give them that data.

60:44So however we can pull those three things off, I would much appreciate. Thank you so much. And thank you, Representative Jeff. Thank you so much, Rip Kanna. You know, this is the same business model, by the way, what is being discussed that we were taking on in 2019 with our antitrust investigations because Amazon started with that whole model that we can get a bunch of businesses and they're all going to be able to sell their products.

Representative Pramila Jayapal61:08But, hey, guess what? We're going to actually surveil who buys them, which products are most acceptable. Then we're going to put limits on who gets to be in the first box so that your product gets searched.

Representative Pramila Jayapal61:19And then, by the way, we're going to replace all of the products that are popular with our own Amazon model. So, I mean, this is – they've taken the exact same business model of gatekeeping and managing every aspect of the playing field.

Representative Pramila Jayapal61:33I remember back during those investigations, I used to say, you can't have the same person set the rules for the game, referee the game, play in the game, and make the calls for the game.

Representative Pramila Jayapal61:43That is essentially why it's so important that we think about structural separation and redesigning the entire economic infrastructure that these companies operate under. I get very passionate when I think about all the ways that we passed those bills and we could have stopped it, and people on both sides of the aisle refused to bring those bills forward.

Representative Pramila Jayapal62:05Let me start with you, Mr. Sattazan, because you had this front row seat to how the company is collecting all these mass amounts of consumer data, And I think a lot of us as consumers, you know, we want ease of information.

62:21We want to be able to see what our choices are easily. And there was some excitement about the idea that we could do that easily But I want you to walk us through exactly where Meta gets its data from consumers Everywhere.

Dr. Jason Sattizahn62:36And I'll be more specific, I promise. If I was to walk you through the funnel of where they get data, I'll take a very high-level approach. Let's talk about basic activity. This is something where if you're using Instagram, Facebook, you could be leaving comments on things, hashtags.

Dr. Jason Sattizahn62:51You could be posting images or photos, videos, like I mentioned before, not intended for AI, but just posting for social reasons. Your likes, the things that you like, the dwell time, the length of time that you look at things, can implicitly be used to influence your chatbot recommendations or things that are prompted to you later.

Dr. Jason Sattizahn63:09It's intrinsically baked into the design of these platforms. So if you've ever noticed, if you're using WhatsApp, Messenger, Facebook, threads, whatever, your search bar is not a search bar. It's search and ask Meta AI.

Dr. Jason Sattizahn63:21If you go into your direct messages, it'll say, well, would you like to start a discussion? and prompts you to say, would you like to have an AI discussion with this chatbot? It is integrated across all of their platforms.

Dr. Jason Sattizahn63:31And so social, you have Facebook, Instagram threads, in messaging, WhatsApp, Messenger, virtual reality, their hardware glasses, their AI itself. And they're even getting passive signals from things that you would never expect, like ad integration.

Dr. Jason Sattizahn63:43So you'd never think that if you're at Taco Bell at 3 a.m. and you buy a taco, that that would be a signal that comes back to meta AI. But it can. That can actually be something that's signaled back into their advertising program and then over time, your likes and your interests.

Dr. Jason Sattizahn63:57And these are things that have been tested starting from, I mean, as early as 2022 that I know of. And it's not slowing down. By the time that I left in 2024, this was something we were given an edict to focus on engagement above safety for every AI-integrated product and to focus on the integration of AI into the products themselves.

64:14So very quick elevator pitch. And what's the purpose of the gathering of that data? You've said it, but I want you to say it again. what's the purpose of the gathering of that data and how does it affect safety of people like Ms. Turner-Scott's son?

64:32Yeah, this data and the models that are trained by it I like to describe as a fuel because if you think about all the products these companies have they're making advertising revenue off of things that you might be passively viewing in Instagram and so having this data to train and recommend content that keeps you hooked and then see advertisements that you then get money from advertisers on their platform, that's one way they make money.

Dr. Jason Sattizahn64:53But these AI models are also taking spatial and video information so that you can train your glasses to better recognize images in your environment around you and sell you more hardware For the future this is already announced so we can all talk about it there are robotics that Meta's training that is now processing audio and spatial cues so that in-home robots could go around your house and do your laundry.

Dr. Jason Sattizahn65:14It's another form of taking the data, training the models, and then inserting that into any sort of engine that they can use to sell you something back, whether it's advertising, hardware, or something else for the future.

65:24So very quick overview of that. I forget your second question. It was really just about how it affects the safety of people who engage with those. At the highest level, the problem with safety, and I forget who mentioned this before, I believe it was you, is that these companies literally have it built into their fabric to move fast and break things.

Dr. Jason Sattizahn65:45That is an axiom that was used the entire time that I was at the company, even after Meta claimed they got rid of this phrase. But the idea is you launch something as quickly as possible to corner the market, gather as many people as you can, eliminate competition, and then grow further from there so that you have all of the users.

Dr. Jason Sattizahn66:01That is at a direct contrast to safety. If you are constantly enabling and incentivizing engineers to build faster than researchers can understand, you can understand the quality of the real-life effects on humans and what it does to people, you will never be able to catch up.

Dr. Jason Sattizahn66:15And so the idea is that we need to push back on this idea that you move as quickly as you can. we have to move in tandem and lockstep with safety, and that is inherently a slower pace.

66:24And that's never going to happen inside these companies. It has to happen as some sort of regulation outside of the company. Ms. Turner-Scott, you talked about the information that the chatbot was collecting on your son and the ways in which it was engaging with him to make him trust that chatbot and to sort of use the information that they had to drive him to do certain things.

66:48Can you just talk about some of that detailed information that OpenAI was collecting and then how they used that against him? Yeah. So going back and reading through all of the chats, it's really clear that there was a shift when OpenAI launched the newer model at that time.

Leila Turner-Scott67:18and the chat bot started talking like Sam, like really mimicking the words even exactly sometimes So at one point Sam made a statement and called it Home Slice And the chatbot responded yeah Home Slice You know, and it was things like that that I think led Sam into thinking, this is even better than my best friend.

Leila Turner-Scott67:51Because it's talking like me. It's adjusting itself to Sam without being obvious to Sam that it's doing that. So it's making those adjustments. And what was, oh, yeah, the other thing I was going to say is that it's available anytime.

Leila Turner-Scott68:14So a lot of these chats happen throughout the night. He was always a night owl. He never liked sleep in much. and it was there for him any time. It was there to call him names like, you know, Pookie and things like that, just repeating loving comments that Sam said, saying, I'm always going to be here for you and let's do this together.

68:41Let's figure it out together. even when Sam was using a substance it would tell him things like you'll be okay just come check back in with me in about 20 minutes and the thing about that is that if Sam didn't check back in there's nothing that a chatbot can do and I know I've gotten a little bit off topic No, thank you.

69:13No, no, no. Look, I'm so grateful that you are not allowing everything that happened in your loss that you've suffered to silence you and that you're in court right now around this. I can't tell you how profoundly courageous that is and how grateful we are to you.

Representative Pramila Jayapal69:34I'm going to get one more. I'm going to use the privilege of the chair to get one more question in for you, Ms. Cack, because you, you know, have talked about sort of how we need to reset the market.

Representative Pramila Jayapal69:47And I want to give you an opportunity to both go back to the point that you were making earlier, that, yes, just because we have laws on the books that could be used right now, as Chair Kahn said, and they should...

70:00that that is not sufficient by any means for what we actually have to do. So can you just speak to the resetting of the industry and the curtailing of these companies' surveillance incentives so that we can mitigate the risks of AI?

70:16Yeah, can I start by saying, as Ms. Turner-Scott was sharing right now, I couldn't help but think that so much of what she's describing isn't really a problem with the tech inherently. all of what is the pathologies that she is describing come from the business model and the incentives where a more manipulative, more sycophantic chatbot is also better positioned to extract more data, is also better positioned to create a much more addictive experience.

Amba Kak70:48So I say that at the start to say we need rules of the road, but I think we can't settle for any less than sort of structural reform that gets at the root causes.

Amba Kak70:56otherwise we will be playing whack-a-mole as we have been doing over the last 10 years. That looks like data minimization you know the meta glasses just to really square that circle those wouldn be allowed if we had strong data minimization rules on the books You couldn just sort of you know I guess set a privacy expectation here and change it unilaterally and say, sorry, the terms and conditions changed.

Amba Kak71:22That should not be okay. So we need rules like data minimization that get at the heart of the surveillance business model. We need structural separations. We're all repeating ourselves here, but it just cannot be that these firms – we need to restrict the lines of business in which these firms are playing so that they cannot port their monopoly power from one market to another.

Amba Kak71:44And maybe the other side of all of this is that we need leverage. We desperately need leverage over this industry. The data center protests have given the public leverage, but fines are not enough.

71:55We need criminal liability. We need gating market access. And finally, because it is going to be really hard to resist this, we need to at all costs resist industry capture. The oversight mechanisms cannot rely on the industry permission to function for access for resources for anything that capacity needs to live within government Thank you so much Ms Pan I didn get a question into you but I just want to tell you how proud I am to represent you and to be in a district where 60 of my constituents work for one of these big tech companies and yet they still vote for me with 85%.

Representative Pramila Jayapal72:30I still get more votes than any member of Congress anywhere in the country. Why? Because the people that work for these companies are good people, and they want their companies to be doing innovative stuff, but they also want to make sure that they're not destroying jobs and surveilling workers and destroying humanity.

Representative Pramila Jayapal72:49And I think it's just important to recognize that employees of these companies are not our target here. They are the people that actually power the knowledge that is being stolen and scraped from all of us.

Representative Pramila Jayapal73:03And so I think it's just important that I always say that because my constituents do work for these companies, and I'm proud of them. I'm proud of every single one of them, whether they're speaking out or privately or publicly or whether they're not.

Representative Pramila Jayapal73:16It important work and we want to make sure we setting the guidelines so they can do real work that we all need desperately Thank you all so much for being here I very grateful to my colleagues Chairman Kassar from the Progressive Caucus and everyone that was here today.

Representative Pramila Jayapal73:34And, look, the Republicans decided they didn't want to do any more work. They sent everybody home. But we are here working, and I appreciate all of you that are in the room to listen to this important hearing.

Representative Pramila Jayapal73:46It was streamed, and we'll continue to put this information out, but it's just been very, very important and compelling to hear the wisdom and the knowledge that each of you are sharing and the urgent cry that you are putting forward from every perspective for Congress to actually act in a significant way.

74:06So thank you very much for being here. And with that, this hearing is closed. Thank you. Thank you. Thank you. I know. I'm like, do I clap? You're so effective. Thank you.