Over the last few weeks, a panic over the possibility of dangerous consequences from artificial intelligence has swept through elite media and politics. I haven’t seen anything like this since the Covid moment, and before that the 2008 financial crisis, the pre-Iraq War debate period, and the few months after 9/11. The fear is thick, and real. And the one solid demand, seemingly from every quarter, is that We Must Regulate This Technology.
The most common solution is to impose some form of safety standards, akin to the Food and Drug Administration, but for large language models. That’s something former Congressional candidate Alex Bores believes, and he’s raised $30 million in just a few months to launch a political organization around it. That’s where Bernie Sanders is, and Anthropic, OpenAI, and Google are seeking something similar. Others analogize the problem to large banks, calling for a bank supervisory regime. Some want a total pause on any AI development.
Many of these ideas are vague and sometimes not administrable, with undefined terms. But there’s nothing wrong with having a basket of ideas. That said, there’s something very weird about this whole situation. And that is, we already have a set of regulators at a Federal and state level with a mandate to look at industrial practices. There are private rights of action where individual citizens and companies can bring lawsuits, and they do. There are also numerous laws in place that already prohibit many of the harmful activities engaged in by the large AI firms. But those laws are mostly not being enforced sufficiently to make a meaningful difference, because in America, we simply do not enforce the law against the powerful. And it’s not clear to me why a new law, an FDA for AI, or even a pause on tech development, would wind up any different.
To understand why, let’s start with the attempts to regulate AI and why they haven’t delivered. The short story is that under the Biden administration, enforcers, most notably Lina Khan at the FTC, were actually responding to safety concerns about AI. Both the Antitrust Division and the FTC brought in a host of technical experts to beef up their capacity. You probably didn’t hear about these moves, but that’s because Joe Biden did not promote or publicize them. Biden was personally uninterested, so were Congressional Democrats. Those that were interested sought to curry favor with big tech, and so kept it quiet. And judges, who are politicians in robes and saw the writing on the wall, tended to side with big tech. Then in 2024, Trump won the election, and his administration canceled and rolled back attempts to enforce the law against the powerful.
Let’s get into specifics. By far the most important and successful action the FTC took was in 2021, when the commission blocked the merger between AI chipmaker Nvidia and Arm. That set the stage for the growth of both companies, who could focus on their lines of business instead of a cumbersome set of turf wars that accompany mergers. Today, for better or worse, Nvidia is the biggest company in the world by market capitalization, the engine of the AI revolution.
There was a lot more that bears directly on safety questions. After OpenAI launched ChatGPT in 2022, the Federal Trade Commission enacted a flurry of studies and investigations looking into the deployment of AI. The commission was building on its work on big tech, which it had been investigating for years. Most notably, in 2023, the FTC launched a probe into ChatGPT, asking very specific questions about OpenAI’s safety practices.
There’s a lot more. The FTC did studies on cross-ownership and acquihires. It brought multiple orders against companies using AI in deceptive ways or building technologies designed to commit fraud. It did work on data breaches, on surveillance pricing, on big tech’s ability to launch new products using machine learning, and even held CEOs personally liable for bad cybersecurity practices. The FTC’s sister enforcers at the Antitrust Division brought multiple monopolization cases against Google, both of which came to involve AI. it also filed a complaint against United Health’s acquisition of Change, a case involving data and machine learning, and it did work on the use of algorithms for price-setting in meat-packing and rent-setting.
So what happened? Well, despite these actions, Khan and Kanter had very little support from Congress. Democratic Senate leader Chuck Schumer’s daughters worked at Facebook and Amazon, and in 2022 he personally blocked antitrust legislation from coming to the floor of the Senate so as to raise more campaign money from big tech donors. Pelosi similarly wouldn’t allow big tech legislation to come to the floor in the House.
When Trump got elected in 2024, the Trump-Vance FTC Chair Andrew Ferguson immediately moved to reverse most of what Khan did, and even tried to erase her entire record, scrubbing the FTC’s website of more than 300 blog posts involving AI.
But it was much more than just symbolic, Ferguson took the unusual step of pardoning an AI offender by setting aside the penalty against, Rytr, an AI development firm, for marketing its AI tools as a way to falsify testimonials and reviews. As powerful white collar defense lawyers noted, Ferguson was “signalling a shift in how the Commission will approach AI enforcement.” He quietly closed a public comment docket on surveillance pricing, and he has presumably ended the investigation into ChatGPT.
These moves are consistent with Trump’s “AI Action Plan,” which called for Ferguson to “review all FTC final orders, consent decrees and injunctions, and, where appropriate, seek to modify or set-aside any that unduly burden AI innovation.” It is also consistent with the Trump Antitrust Division’s approach to algorithmic price-fixing, which it basically endorsed by settling the RealPage and Agri-Stats meat price-fixing cases.
Not enforcing the law against the powerful is pretty much administration policy. On Monday, Attorney General Todd Blanche announced explicitly that the Department of Justice lean strongly against litigating against AI companies. “The last administration spent countless prosecutors’ hours and money and effort and resources regulating whatever they chose to regulate,” Blanche said. “I think that there are some Justice Departments that like to do that and that’s not what we’re doing.”
While there are partisan valences here, there has also been a broader elite consensus in the judiciary that trying to impose meaningful legal obligations on large technology and AI firms is ridiculous. In 2023, three important D.C. Circuit judges dismissed government anti-monopoly claims against Meta. They said the case was simply “odd,” as the very premise of litigating market power violations against an innovative high-tech enterprise was foolish.
Last year, Judge Jeb Boasberg ruled that Meta was not a monopoly. Judge Amit Mehta ruled that though Google was an illegal monopoly, he would not impose any meaningful remedy. Then two days ago, Judge Leonie Brinkema unsealed her opinion in another Google case, writing in somewhat mean-spirited tones that yes, Google was an illegal monopoly, but she would impose no meaningful remedy. She even indicated there were simply no circumstances in which she could ever imagine breaking up a monopoly. Boasberg, Mehta, and Brinkema were all Democratic appointees.
So why go through this history? Well, it’s because I am trying to show the problem is not a lack of AI regulation. We have AI regulation. The problem is there’s an elite consensus that the rule of law simply does not apply to the powerful.
Take the law in an entirely different context. It is per se illegal to engage in a price-fixing conspiracy, what is in the law as a prohibition against a “restraint of trade.” But right now, billionaire Robert Kraft is openly organizing a conspiracy with other stadium and venue owners to deny access to music venues against the artist Macklemore because of his song protesting the genocide in Gaza. That is a potential criminal or illegal act, and not only is it going without litigation or prosecution but it’s not even considered that Kraft should be held liable for what he’s doing. He’s a billionaire, he owns Gillett Stadium, and that’s that. There’s cultural pushback, but the concept of legal pushback feels comical. There’s no happy ending in this story.
There is an endless parade of this kind of lawlessness. It was reported years ago CVS would lower reimbursement rates to pharmacies to kill their business, and then offer to buy them out. On a local level, I just heard a story about a city in Colorado where police and law enforcers simply would not enforce laws prohibiting landlords from engaging in certain practices. They simply refused, because landlords are rich.
John Deere dealers can threaten farmers who raise a fuss, United Health can scare dying patients into not requesting the care they need, and Paramount can threaten the entire state of California if it is not allowed to break the law. We no longer live in a land where people are “innocent until proven guilty,” the actual rule of law as practiced is “guilty until proven wealthy.” The Supreme Court even gave Wall Street an exemption from the Constitution in securing bankers their own independent regulator, even though no one else gets a regulator free from Presidential demands. Oh, and let’s not even talk about laws that purport to constrain illegal wars or war crimes.
And this dynamic is widely understood among normal people. In a recent NBC poll, 54% of Americans agreed that “When it comes to politics and society, nothing really matters because powerful people will always do whatever they want.”
So let’s get back to the problem of AI, and the recent panic. It’s not that hard to make this technology safe, as the CEO of Hugging Face indicates. It requires some prudent risk management, some liability for big AI firms, and beefed up cybersecurity investment and requirements for corporations.
Still, I’ve had subscribers cancel from BIG because I have not joined in the hysteria, and former allies tell me I’m doing the equivalent of pushing Ivermectin during Covid by saying that Anthropic should stop its IPO. And I think there’s a reason this fear is so potent, and why it’s so hard to think calmly and rationally about what to do. And that is because it’s almost painful to imagine a world where we enforce laws as written against the powerful. As a society, we have lost the ability to protect ourselves, and that is a very scary thing to experience.
This inability has clouded our judgment as to what we are facing. Indeed, most people imagine AI companies as just ordinary firms that happened upon a groundbreaking technology. But what they are is the result of an unprecedented crime spree.
It’s not just that the hyperscalers are illegal monopolists, or that Sam Bankman Fried was the most important initial funder of Anthropic, or that Meta’s business was caught for mass sex trafficking of children, or that there’s a huge amount of financial chicanery involved in funding the data center buildout.
It’s much more direct; in unsealed legal documents discovered by Jason Kint in a case brought by the New York Times over copyright violations by giant AI firms, OpenAI admitted circumventing paywalls to scrape content. The company’s President Greg Brockman, when told his company had hacked the NYT to scrape the site, responded with "ah nice." And in those same documents, it came out that Microsoft’s Director of Applied Science called the training of big AI models on copyrighted content the “largest theft of labor in human history.” These actions may be a violation of the Computer Fraud and Abuse Act, which prohibits hacking into computer systems and taking things of value. It could be a criminal violation of copyright law. But at some level, the “largest theft of labor in human history” must be against some criminal law.
Just imagine if our enforcers and judges took that rhetoric seriously. If we enforced the law against the powerful, if we stopped this theft, it would radically upend how society works. It would let us feel a sense of control once again. And yes, it’s quite possible to apply written laws. Certainly, if you could charge Aaron Swartz, a genius programmer hounded to death for accessing JSTOR articles by a prosecutor in 2013, you could charge OpenAI.
On Monday, former Khan wryly made that general point, when she posted a statement discussing laws that are already on the books that could be used to address the problem of unsafe AI products. These include consumer protection laws, laws against unfair and deceptive conduct, cybersecurity requirements, and so forth.
Khan mentioned that state enforcers are looking at potential criminal liability against AI firms. And yes, they are investigating, even if the Trump administration isn’t.
After I published Khan’s arguments, I got an angry email from a reader, arguing I’m not taking the AI doomsday problem seriously.
The idea that “law and order” will prevail -- do you see any evidence of that in real life at the moment? -- or that government will be competent to step in with meaningful regulation -- when many of our reps don’t know how to use their phones -- is absurd. Not to mention, which government?
She then approvingly cited this piece, by Stephen Witt, titled “This Is Really Bad,” in the New York Times. Witt described his fears about this uncontrollable technology, and then suggested we should pass laws banning AI development, transparency, kill switches, and investigations/regulation. That, she argued, was a “constructive serious agenda.” Her view is quite common, I’ve had many conversations among politicians and activists who feel similarly. Something Must Be Done, and That Something Must Be Big and Important.
What is odd is not so much the desire for action, but a core contradiction here. If enforcing current laws against powerful people is impossible, why would more law help? What exactly is more regulation going to get you that existing regulation doesn’t, if none of it will be enforced? Demanding action by government while sneering at the possibility of government action is incoherent.
Usually, when there’s something this obviously contradictory in the core thinking of a large swath of political elites, the actual problem is not what’s being debated. And in this case, I think what’s happening is that there’s a deep feeling of nihilism resulting from what we all know, which is that the rule of law as written does not reflect the rule of law as applied. Trump just openly dismissed the idea of law as a guiding principle of our social order, and that has severely damaged our faith in it. Katy Perry, for instance, proudly displayed her Anthropic subscription, with a heart drawn on a screenshot when Pete Hegseth tried to call the company a supply chain risk.
There is winnowing confidence in the ability of political actors to enforce laws fairly. Again, this will be no different if we pass new laws, and evisceration of the administrative state is reflected in reduced confidence in the expertise and independence of federal enforcers. Private rights of action, which are generally disliked by politicians, are a potential path, as are state enforcers. But we need a bigger cultural and political shift, a recognition that equality before the law is a fundamentally radical project, and it’s one we must fight to achieve.
That’s a hard case to make. The rule of law when mouthed by self-satisfied liberal politicians, sounds problematic in two ways. First, arguing for the rule of law sounds like you support the unsatisfying status quo. Paradoxically, it also sounds unrealistic. The idea of using the law to put Sam Altman on trial, well, that sounds like a utopian fantasy more unrealistic than bots destroying the world.
In other words, the very notion of arguing for applying the rules as written in a reasonably equal manner, the very basis of a written Constitution animating the American project, seemingly means you’re an out of touch elite or you’re a delusional romantic. And this collective desire for a vague “regulation of AI,” or an “FDA for AI,” or to “pause AI development,” reflects a hunger for a deus ex machina device to get around what is in effect a Constitutional crisis.
That’s not to say I would oppose new rules or regulatory agencies. It’s just that these proposals strike me as besides the point. Maybe it’s necessary to have a level of panic and new laws to spur action, but I just think it’s important to recognize that the elite consensus that rules don’t apply to the powerful simply cannot coexist with any credible attempt to make a harmful technology more safe. It is that consensus that blocks the existing rules from taking effect, and that will block any future regulation from being effective at making these technologies safe.
After all, in an alternative history in which that elite consensus protecting Sam Altman didn’t exist, and the regulatory actions of the last four years were coming to fruition, as Google was selling off its component pieces in response to a harsh antitrust remedy, then OpenAI’s new CEO, who took office when Sam Altman was indicted, would be engaged in a crash safety program to make AI systems safe, as would the entire industry. But laws to force that are already on the books. It’s just that we know what’s written down isn’t what matters.
Thanks for reading! Your tips make this newsletter what it is, so please send tips on weird monopolies, stories I’ve missed, or other thoughts. And if you liked this issue of BIG, you can sign up here for more issues, a newsletter on how to restore fair commerce, innovation, and democracy. Consider becoming a paying subscriber to support this work, or if you are a paying subscriber, giving a gift subscription to a friend, colleague, or family member. If you really liked it, read my book, Goliath: The 100-Year War Between Monopoly Power and Democracy.
cheers,
Matt Stoller