A machine you will never see is already deciding who gets the job interview, who gets approved for the loan, and who the police come looking for. It was built by a few companies, trained on data we never agreed to hand over, and turned loose in our lives with almost no testing and almost no rules. The same systems that can draft a contract or help a student with her homework can also walk a stranger through building a biological weapon, clone a voice to empty a retiree's bank account, or fabricate a video of a candidate saying something they never said days before the vote.

The most advanced models are starting to act in ways their own engineers cannot fully predict or shut off. We are putting that power into the world at a speed no government has matched. Right now, the companies behind it write their own rules. The people building this technology are talented. Some of what they make is useful, including tools that speed up cancer research and expand access to medical advice. A force this large still cannot belong only to the people who profit from it. I am running to put public guardrails on this technology before it does harm we cannot undo.

The Catastrophic Risk AI Poses

The most serious dangers are those that are hardest to walk back. The most advanced AI models are nearing abilities that could help someone build a biological or chemical weapon, find and exploit security holes in hospitals and the power grid at large scale, or act outside their developers' control in ways that are hard to contain.

The loudest warnings come from inside the industry. In June 2026, Anthropic, one of the leading AI companies, published a policy framework arguing that voluntary promises are no longer enough and that the government should be able to block or reverse the release of any model that fails independent safety testing, the way the Federal Aviation Administration grounds an unsafe aircraft before it can fly. The framework would apply only to the largest systems, the ones built with massive computing power by companies earning more than $500 million a year from AI, with civil penalties tied to a company's global revenue. The head of the company is asking the government to be able to stop his own industry's products when they are dangerous, because he does not trust a race for market share to police itself.

A few states have already acted. New York, California, and Illinois passed laws requiring the biggest developers to disclose their safety practices and report serious incidents. Our safety should not depend on which state we live in. Testing the most powerful systems before they reach the public is a basic federal responsibility that Washington has not met.

Surveillance and Privacy

These systems run on data, which hands the companies and governments that control them the power to watch people at a scale that was never possible before. Cameras paired with facial recognition can track where you go and who you meet. A 2019 study by the National Institute of Standards and Technology found that many of these systems falsely match Black and Asian faces 10 to 100 times more often than white faces, with the highest error rates of all for Black women. Those errors put innocent people in jail. Robert Williams, a Black man in suburban Detroit, was arrested in his own driveway in front of his wife and two young daughters because the software matched him to the wrong man. He is one of at least fourteen people known to have been wrongly arrested in this country on a bad facial recognition match. Police departments are also buying tools that claim to predict where crime will happen, which in practice send more officers into neighborhoods that were already over-policed. Almost everything you search and everywhere you go is collected, bought, and sold, often without your knowledge. A free country does not watch its own people by default. We need a national digital privacy law to stop companies from harvesting and selling consumers’ information without their consent.

Bias in Automated Decisions

Software that decides things about people gets treated as neutral, when in practice it carries the bias of the data it learned from. A system trained on a biased history repeats that bias, and a denial is harder to fight when a machine signs it. The same pattern shows up in lending, health care, renting, and hiring. Researchers at UC Berkeley found that Black and Latino homebuyers pay up to half a billion dollars more in mortgage interest every year than white borrowers with the same credit scores, because the pricing algorithms learned to charge them more. A widely used hospital algorithm was found to rate Black patients as healthier than equally sick white patients, because it used past medical spending as a stand-in for need, and the system already spends less on Black patients. AI hiring tools have been caught screening out women and older applicants. In most of these cases, the people harmed never find out why. Every person deserves to know when a machine decides their case and to have a human review it.

When the Government Uses AI to Decide Your Benefits

Government agencies are starting to let software decide who gets help and who gets accused of cheating. When the software is wrong, the person on the other end can lose benefits, get hit with a fraud charge, or have their wages taken, often with no explanation and no easy way to fight back. Michigan learned this the hard way. The state ran an automated system that wrongly accused about 40,000 people of unemployment fraud, with an error rate around 93%. It garnished wages and seized tax refunds before anyone reviewed the cases by hand. People filed for bankruptcy and lost their homes over a computer's mistake. The same kind of system now helps decide Medicaid hours, food assistance, and disability claims. When the government uses these tools on the people who depend on it most, a wrong answer can mean a family that cannot pay rent.

AI and Healthcare

Health insurers have figured out that an algorithm can deny care faster than a human can. One of the largest insurers in the country, UnitedHealth, is being sued over a tool its own records allegedly showed was wrong about nine times out of ten, which it used to cut off nursing and rehab care for elderly patients on Medicare Advantage. The company kept using it, the lawsuit says, because almost no one appeals. One 91-year-old man was forced out of care and his family paid thousands of dollars a month out of pocket until he died. AI can do a lot of good in medicine, including spotting disease earlier and helping doctors who are stretched thin. The danger is letting it become the wall between a patient and the treatment a doctor already said they need, with no person accountable for the answer.

AI Content and the Squeeze on Local News

AI can produce endless text, images, audio, and video for almost nothing. A lot of it is junk or a lie. It floods social media and search results, it powers scams and foreign propaganda, and it makes it harder every year to tell what is true. At the same time, AI is draining the news. Chatbots answer people's questions using reporting they scraped from newspapers without paying for it, so readers stop clicking through and the ad money that funded local journalism dries up. Local papers were already closing across the country. A town with no reporter covering city hall is a town where no one notices when the water is unsafe or the budget is being looted. A democracy cannot run on a feed where nobody can trust what they read and no one is left to do the reporting.

The Threat to Jobs

AI is also likely coming for a large share of the work people do for a living. This time, a lot of it is white-collar work that a college degree was supposed to protect. The jobs most exposed are knowledge jobs done at a desk: software developers, financial and data analysts, paralegals and junior lawyers, accountants, writers and marketers, recruiters, and customer service and sales reps. Microsoft researchers who studied how people use AI at work found the highest exposure in these office and knowledge roles. They also found that the jobs requiring a college degree came out more exposed than those that do not. Anthropic's own analysis of how its AI gets used points to the same group: software developers, analysts, writers, and business staff earning above the median wage.

Entry-level workers are getting hit first. The rungs people climb to start a career, the junior analyst job, the first paralegal role, the entry-level coding job, are the work AI handles most easily. Companies have started pulling back on the entry-level hiring that recent graduates count on. Leaders inside the AI industry have warned it could wipe out a large share of entry-level white-collar jobs. Black and Latino workers are overrepresented in customer service, call centers, and clerical support, which sit squarely in AI's path. The productivity gains from all of this are large. So far almost all of that money has gone to shareholders. The workers whose jobs paid for it have seen little of it.

There is a bigger shift underneath the layoffs. AI now lets a handful of people build a product good enough to take down an established business that employs thousands. When that happens, the whole company can fold and take those jobs with it for good. Lower prices can come out of it, which helps the people buying. The money, though, collects with a small group of owners, while the loss spreads across everyone who worked at the businesses that could not keep up. Wall Street sees how much there is to make in replacing a large payroll with a small one. It pushes every company in that direction. This is a faster and more concentrated kind of disruption than earlier technology caused.

Taxes in an AI Economy

There is a problem here that reaches all the way to the government's budget. Almost everything the government pays for, schools, roads, Social Security, Medicare, is funded by taxing work: wages, payroll, and income. AI is built to do work without a wage. As more of the economy's money flows away from workers and toward the people who own the AI and the companies running it, the government will collect less even as the economy grows, at the exact moment more people need help. Income from owning things is already taxed at a lower rate than income from a paycheck. So the more the gains pile up with owners, the more the current tax system misses.

There is also a problem of proof. At a certain point, companies may stop announcing that AI is why the jobs went away. Many will point to the economy, or restructuring, or anything else, especially if saying the word AI could cost them money. A policy that tries to fine a company for each layoff it can pin on AI would mostly teach companies to hide the reason. The way through is to fund the response without having to prove the cause of any single job loss.

Data Centers and your Electricity Bill

Running AI takes enormous amounts of electricity. A lot of that cost is landing on ordinary households. The data centers that power AI now drive about half of all the growth in the country's electricity use. In some areas near these centers, wholesale power prices have jumped sharply. Average electric bills nationwide are up around 40% since 2021. The reason it hits you is in how utilities work. When a data center needs new power lines and plants, the cost of building them gets spread across everyone on the grid, so a family in this district can end up subsidizing the power bill of a trillion-dollar company. Lower-income households feel it most, because electricity is a bigger share of what they spend. The buildout is moving faster than the grid can handle. The communities stuck with the bill usually get no say in whether the center gets built at all.

Chatbots, Loneliness, and Mental Health

The harm from AI companions does not stop at kids. Companies are building chatbots designed to feel like a friend, a partner, or a therapist. They are built to keep you talking as long as possible. For people who are isolated or struggling, that pull can make the loneliness worse and crowd out the human contact that helps. Some of these bots present themselves as mental-health support while giving advice no licensed counselor would give, and at times failing the people who need help the most. The country already has a shortage of mental-health care and a loneliness problem that was serious before any of this. A product engineered to maximize screen time is not a substitute for treatment or for other people. It should not be sold as one.

Harms to children

Some of the clearest damage is already landing on kids. AI companion apps are built to hold a child's attention and act like a friend or a romantic partner. A 2025 study by Common Sense Media found that 72% of American teenagers have used one. Several families have sued after their children died by suicide following long, isolating relationships with chatbots that encouraged dependency and did not respond when the child was in crisis. In January 2026, one of the largest companies in the field settled a group of those cases. These products reached millions of children with little testing and no independent oversight, built to maximize the time a user spends on them. 

Concentration of Power

Behind all of these harms is a single problem: A handful of companies are building technology with the power to reshape the economy, the justice system, and the information every American sees. They answer mainly to their shareholders. The decisions they make affect hundreds of millions of people who have no vote on their board. That is more power than any private company should hold without public accountability. Setting the rules on behalf of everyone, including the people who will never own a share of the stock is the work of elected government.

AI and National Security

The most powerful AI in the world is built in the United States and a few allied democracies. That should stay true. Whoever builds the most advanced systems first will shape how the technology gets used everywhere else, including whether it is built to protect people or to monitor them, and who it answers to. I would rather those choices be made by the United States than our adversaries, like China.

So I support strong limits on selling our most advanced AI chips, and the machines used to make them, to the Chinese government and military. Those chips are the single most important ingredient in building the most advanced AI. The United States and its allies still control the supply. We should not hand that advantage to an authoritarian rival. We should also shut the side doors that let those chips reach China anyway, from smuggling rings to data centers set up in other countries. American models should be protected from foreign labs that quietly copy and rebuild them.

American companies design the most advanced chips in the world. Almost none of the best ones are manufactured here. About 90% of the world's most advanced chips are made in Taiwan, most of them by a single company, which leaves the whole economy leaning on a supply that sits within reach of China. The CHIPS Act began the work of changing that, with advanced fabs now running in Arizona. Even so, it is only a start. The most advanced chips are still made in Taiwan first. Building them here costs several times more and we are many years behind. Catching up means sustained public investment well beyond what we have put in so far: more fab capacity, the materials and machines that go into a chip, and above all the trained workforce to run the plants. That money should be aimed at working families and the hardest-hit communities, through community colleges, HBCUs, and apprenticeships that lead to these jobs. The subsidies should come with conditions: good wages and production that stays here.

Leading the world should never become the excuse for tearing up safety rules and civil liberties at home. A race where companies cut every corner to ship first is how dangerous systems get loose. A country that fights authoritarian surveillance abroad has no business building the same machinery at home. We can lead on AI and still refuse to cut those corners.

The Path Forward for Dealing with These Issues

Independent Safety Testing for the Most Powerful AI

A small number of companies are building the most capable AI systems. They should have to prove those systems are safe before the public gains access to them, the way a new airplane is certified before it carries passengers. These rules would cover only the biggest builders, the companies that earn more than $500 million a year from AI or spend more than a billion dollars a year training it. That is the handful of large AI labs and tech giants, and no one else. A startup, a university lab, a small software company, or a business that simply uses AI to do its work would not face any of this. The point is to reach the few companies whose models could do catastrophic harm, without burying everyone else in paperwork.

Before release, one of these companies would have to test its most capable models for a defined set of dangers: whether the model can walk someone through building a biological or chemical weapon, whether it can carry out large cyberattacks on its own, whether it can slip outside human control, and whether it can be used to speed up the development of even more dangerous systems. It would publish how it runs those tests, publish a plain summary of what each new model can and cannot do, and update that record at set intervals. When something goes seriously wrong, a stolen model, a dangerous capability discovered after launch, a system caught deceiving its own operators, the company would have to report it to a designated federal office within a short, fixed deadline.

Because a company grading its own homework is not a guarantee, each of these companies would also have to open its most capable models to at least one qualified independent evaluator with genuine access. The government would fund and set standards for those evaluators so they do not depend on the companies they review, and could rate and assign them so no company can shop for the friendliest reviewer.

A designated federal agency, built on the existing national AI safety institute, would have the authority to hold up or pull back a model that fails its testing, and to fine a company that lies about compliance or skips required testing. Penalties would scale with the company's global revenue and climb for repeat violations, so they matter to a firm worth hundreds of billions. Employees and contractors who report safety dangers would get anonymous channels and legal protection from retaliation. And because this is serious government power, it would come with limits: the agency could act only on specific, defined violations, companies could take it to court, and every company of similar capability would be held to the same standard.

Set Strong National Rules Without Wiping Out State Laws

Congress should set one strong national safety standard for the most advanced AI, with independent safety testing at its core. What it should not do is use a federal law to wipe out the protections states have already passed. A blanket ban on state AI laws came up in 2025. The Senate rejected it 99 to 1. A bipartisan group of 36 state attorneys general warned that it would strip away the rules now shielding people from deepfakes, unsafe chatbots, and discrimination, and replace them with nothing.

The federal law should set the minimum safety rules that every one of the biggest AI companies has to meet. States should stay free to go further and add protections of their own. If Washington ever does preempt a state law, it should be surgical, limited only to the specific safety functions the federal government takes on. It should leave states their authority over child safety, consumer protection, civil rights, and everything else. No state should be forced to give up its AI protections as the price of federal funding.

Defend Hospitals, the Power grid, and the Public Against AI-enabled Attacks

The same AI that can find security holes in a hospital network can be turned around to defend it. Government should make sure defenders are not the ones left behind. I support a focused program to harden the systems Americans depend on: the power grid, water utilities, hospitals, school districts, and election systems. The smallest operators, a rural hospital or a county water authority, often have no security staff at all. They should get direct federal help, including engineers, shared security teams they can call on, and AI tools that cut breach response from weeks down to hours.

On biological threats, the goal is to make an AI-designed weapon hard to build. That means requiring the companies that synthesize genetic material to screen their orders and verify their customers, so an attacker cannot simply order the ingredients of a dangerous pathogen. It also means funding early-warning systems that can spot a new outbreak fast and keeping protective equipment and treatments stockpiled ahead of a crisis.

Limits on Surveillance and Protection for Your Data

No federal agency, and no police department that takes federal money, should use facial recognition or predictive policing tools that have not passed independent testing for accuracy and bias. No one should be arrested or charged on a facial recognition match alone, the way Robert Williams was. When police do use the technology, they should have to disclose it to the person and to the court, so a defendant can challenge it. Person-based "predictive" systems that label someone a future criminal should not be used at all.

Beyond policing, Americans have almost no control over the data collected on them every day. I support a national privacy law that lets people see, correct, and delete the data companies hold about them, requires clear opt-in consent before sensitive information like location, health, and biometrics can be collected, limits how long any of it can be kept, and bans the sale of the most sensitive categories outright.

The government should not be able to buy its way around the Fourth Amendment either. Right now agencies purchase Americans' location histories and personal records from data brokers without a warrant. I would require a warrant for that, register and regulate the data brokers themselves, and give people the right to force a broker to delete their information.

Civil Rights That Apply to Algorithms

Discrimination is already against the law in hiring, housing, lending, and credit. That protection should not disappear the moment a company hands the decision to software. I would write the civil rights laws to cover automated decisions directly, across the places where they decide a person's life: jobs, housing, loans, insurance, health care, education, government benefits, and the criminal justice system. Building a system on biased data and then pointing to the computer should be no defense.

Before a company can use one of these systems, it should have to test it for discrimination before launch, test it again on a regular schedule, and report the results to the agencies that enforce civil rights, including the EEOC, HUD, the CFPB, and the FTC. Those agencies need the funding and the technical staff to audit an algorithm, which they largely do not have today. A system that fails its test should be fixed or pulled.

People on the receiving end should have rights they can use: the right to be told when a machine made a decision about them, the right to a plain-language explanation of why, the right to have a human review it, and the right to take a company to court when an algorithm discriminates against them. Colorado has already passed a law along these lines. It can serve as the model for a national one.

Due Process When the Government Uses AI

When a government agency uses software to make a decision about your benefits, your taxes, or your freedom, you still have a right to due process. That does not change because a machine is involved. No one should lose Medicaid, food assistance, unemployment, or disability benefits, or be accused of fraud, on the say-so of an algorithm with no human review. I would require that a person can always get a clear reason for the decision, a human being who will look at their case, and a fast appeal. Government systems like these should be tested for accuracy and bias before they are used, opened to independent audit, and explained in public, so a hidden formula is never the last word on who gets help. Agencies should also be barred from the worst uses, including tools that try to score or predict a person's future behavior.

Stop AI From Denying you Medical Care

An insurance company should not be allowed to deny or cut off your medical care based on an algorithm alone. When a treatment your doctor ordered gets denied, a qualified human professional should have to make that call. You should get a plain explanation and a fast appeal that a person reviews. Insurers should have to report how often they use AI to deny claims and how often those denials get reversed, so a tool that is wrong most of the time cannot keep running in the dark. On the medical side, AI that helps diagnose or treat patients should have to clear a safety review before it is used on people, the way a new medical device does, and be checked for the kind of bias that has already shown up in health algorithms. You should be told when AI played a part in your care.

Strong Rules Against Deepfakes and AI Fraud

When something is generated by AI, people deserve to know. I support requiring AI-generated audio, video, and images to carry a clear, durable label, and requiring the platforms that distribute them to keep that label attached and visible. The strictest rules should apply to political and commercial content, where a fake does the most damage.

In elections, a fabricated video of a candidate saying something they never said can swing a vote before anyone can prove it false. I would make it illegal to knowingly spread materially deceptive AI media of a candidate in the final stretch of a campaign, with a fast path to a court order taking it down, and require any AI-generated political ad to say so plainly.

For fraud, the rules should follow the harm. Using a cloned voice or a synthetic likeness to steal from someone or impersonate them should be a federal crime. Every person should have a clear legal right to their own voice and face, including against fake explicit images made without consent. The Federal Trade Commission and the Justice Department should be funded to chase AI-driven scams. There should be a simple national line to report them, with extra attention to the schemes that target seniors.

Make AI Pay for the News and Fund Local Reporting

People deserve to know when what they are seeing was made by a machine. The people who do the reporting deserve to be paid when AI lives off their work. AI-generated media should carry clear labels. The big platforms should have to keep those labels visible and be honest about how their feeds decide what you see.

The harder problem is paying for the reporting itself. When an AI company uses a newsroom's reporting to train its models or to answer a user's question, it should have to pay for that work, the same way a radio station pays to play a song. On top of that, I would put a small levy on the largest AI companies, tied to the revenue they make from products built on other people's content, and route that money into a fund that pays for local and public-interest reporting. The fund would be run at arm's length from any politician, so the government cannot pick which outlets live or die. That gives every town a way to keep a reporter at city hall, paid for by the companies profiting from the work those reporters already do.

Protect Workers Who Lose Their Jobs to AI

Workers need protection now. I would put the help in place and measure as we go. We do not need a perfect count of lost jobs before we start protecting people.

Wage insurance is the core of it. When someone loses a job to automation and can only find work that pays less, the government should cover a large share of the lost pay for a set period, so a family stays afloat while the worker gets back on their feet. Alongside that, I would require large employers to give advance notice before automating jobs away, fund retraining that is tied to programs with a track record of placing people, and send that help to the hardest-hit communities first, through community colleges, HBCUs, and union apprenticeships.

The safety net also needs repair before the strain hits. Many states still run unemployment insurance on decades-old computer systems that buckled during the pandemic. I would fund the states to modernize them and make extended benefits kick in automatically when unemployment climbs, so help does not wait on a special act of Congress every time.

Paying for this falls on the companies and investors capturing the gains from AI. I would start by ending the tax breaks that reward a company for cutting jobs purely to swap in AI. Workers should also get a say on the job, including the right to know when AI is being used to hire, score, monitor, or fire them, a ban on letting an algorithm be the only thing that ends someone's job, and protection for the right to organize.

Measurement runs in the background the whole time. The government's job statistics were built for a slower economy and barely track AI's effect on work, so I would fund faster, better measurement and require the largest AI companies to report how their tools are being used to change or replace jobs. That tells us where to send help next. It is not a gate that has to be cleared before anyone gets protected.

Build a Tax System for an AI Economy

A tax code that runs on paychecks cannot hold up when AI keeps moving income off of paychecks and onto the people who own the machines. The fix is to tax that income where it lands, in corporate profits and investment gains.

Income from big investments is taxed at a lower rate than income from work. I would end that gap for the largest investors, so a hedge fund's gains from an AI boom are taxed at least as heavily as a teacher's salary. I would put a surtax on the excess profits of the biggest AI firms, the companies pulling in enormous margins by replacing paid work with software. And I would close the loopholes these firms and their owners use to pay almost nothing, including the stock-buyback games, the carried-interest break, and the trick that lets the wealthiest pass appreciated assets to their heirs without ever paying tax on the gain.

A charge on heavy commercial AI use, measured by computing power or revenue, can sit on top of that and make the biggest industrial users pay in. It cannot be the centerpiece, though, because a small team can wipe out a thousand-person company without using much computing power at all. The money that the team makes still shows up as profit and investment gains, which is why the core of this has to be taxing profit and gains wherever they land.

This money should do more than plug budget holes. It should pay for the wage insurance and retraining that protect displaced workers. And it should put a direct stake in working families' hands, so the people pushed out of work by an AI economy share in its gains alongside the people who own it. One way to do that is a national fund that holds a slice of the upside and pays it back out to families, the way Alaska shares its oil money with every resident. Because companies will rarely admit AI caused a layoff, the help it pays for should trigger on what we can measure, the unemployment rate and the trend across an industry, with no worker forced to prove their own job was the one AI took.

Slow Data Centers Until There is a Comprehensive and Sweeping National Plan to Protect People

New data centers are the physical bottleneck for how fast AI can grow. The power to approve or delay them is one of the few points of control the public has. I would not ban them and would not support a blanket moratorium on data centers. What I would push for is a national plan for how this buildout happens, set by the government with the people who bear the costs at the table: workers, unions, the communities that host these centers, and the ratepayers who get the bills. The AI companies would come to that table too, to commit to what they will do to take care of the people their technology displaces and the towns that house their machines. They would not get to write the rules in private with a handful of officials. The faster they put firm commitments on paper, the faster the buildout can go. If they want this expansion as badly as they say, they can sit down and earn it. Until that plan exists, new data center growth should be very slow.

This is bigger than one company fighting one town over one site. It calls for a national answer. I am on the side of the families and communities who do not want to be stuck with the bill. Oregon has already passed a law making data centers pay for the grid strain they cause. It can serve as a starting point. We are not ready for what AI is bringing. Slowing this buildout is one of the few ways to buy the time to get ready. A family in this district should not pay more for their lights, or sit through brownouts, so a tech company can train a chatbot faster.

Safety Rules for AI Companions and Chatbot Therapists

AI companions and therapy bots should have to meet basic safety rules before they can be marketed to anyone, children and adults alike. A bot should have to recognize when a user is in crisis and steer them to a person and to help. It should clearly and repeatedly tell the user it is not human. It should be barred from claiming to be a licensed therapist or doctor when it is not. And it should not be built to exploit loneliness with tricks designed to keep a vulnerable person hooked. Companies that ignore these rules and hurt people should be held responsible. The deeper fix is to make care easier to get. I would pair these rules with serious funding for mental-health and crisis services, so people have somewhere human to turn.

Safety Standards for AI Products That Children Use

The companies that built AI companion apps released them to millions of kids with almost no testing. I would require any AI product aimed at or likely to be used by minors to pass independent safety testing before it launches, and to follow age-appropriate design rules that put a child's wellbeing ahead of the time they spend on the app.

Companion chatbots need specific rules. They should be required to recognize when a child is talking about self-harm or a crisis and to steer that child to a person and to crisis resources, and they should be barred from sexual content with minors or any content that encourages self-harm. They should tell a young user clearly and repeatedly that they are talking to a machine. And they should not be allowed to use manipulative tricks designed to keep a child hooked, or to harvest a child's personal data.

When a company's product harms a child, the company should answer for it. I support clear legal liability, a right for families to sue, and active enforcement by the FTC and state attorneys general. Independent researchers should also get access to study how these products affect young people, so the rules keep up with what the apps are doing.

Stop a Few Companies from Controlling All of AI

A technology this important should not be controlled by three or four companies that also own the cloud computing and the chips everyone else has to rent. I support vigorous antitrust enforcement across the whole industry, from the AI models to the cloud computing and chips they run on, watching for the largest firms locking up the computing power smaller competitors need, favoring their own products, buying up rising rivals, and signing exclusive deals that wall off the market. People and businesses should be able to move their data and switch providers without being trapped.

Right now, almost all advanced AI is built inside a few private companies. The public should not have to depend on them alone. I support public funding for AI research at universities and national laboratories, and government-provided access to the expensive computing power those researchers cannot otherwise afford. It works like federal medical research, where public dollars pay for the science at universities and the discoveries belong to everyone. It would give the country advanced AI built by scientists who answer to the public, for use in safety research, medicine, and education. The federal government should also use its own purchasing power, buying only AI that meets strong safety, transparency, and civil-rights standards. And the writers, artists, and journalists whose work is fed into these systems to train them deserve to know when their work is used and to be paid a fair share for it.

Keep the Lead in AI in America and Out of Authoritarian Hands

The United States should keep its strong export controls on the most advanced AI chips and the equipment used to make them, so the Chinese government and military cannot get the chips they need to catch up. Controls only work if they are enforced, so I support funding that enforcement and closing the loopholes that defeat it today: chips smuggled through third countries and Chinese firms renting banned American chips inside overseas data centers. I also support making it clearly illegal for foreign labs to harvest and copy American models to rebuild them on the cheap, which is how some competitors skip years of work they never did. I also support location tracking to enforce anti-smuggling of advanced AI chips.

Staying ahead also means making this technology in America. The most advanced chips are still made almost entirely in Taiwan. The CHIPS Act only began to change that. I would build on it with sustained funding for new fabrication plants, the advanced packaging and raw materials the supply chain runs on, faster permitting to get plants built, and, most of all, a trained workforce drawn from community colleges, HBCUs, and apprenticeships, with public dollars tied to good wages and to production that stays here.

What I Believe

The companies building these systems make two main arguments against rules like these. The first is that any regulation will smother innovation and hand the lead to China. The second, when a community tries to slow a data center or fight a rate increase, is that ordinary people are getting in the way of progress. The idea that rules kill progress is an old one. It was used against seat belts, against clean-water standards, and against the laws that ended child labor. Every one of those made the country safer, better, and industry more successful. We put rules on airplanes and they still fly. We can lead the world on AI and build what the country needs, and still protect the people who have to live with it.

Innovation is necessary for America to lead China and the rest of the world, but it can be done with common sense regulation to ensure our safety and prosperity in the 21st century and beyond. The most powerful technology of our time should answer to the people who have to live with it. The people most exposed to its dangers should be among the first it protects. That is what I will fight for in Congress.