Category: Law
AI, Unemployment and Work
Imagine I told you that AI was going to create a 40% unemployment rate. Sounds bad, right? Catastrophic even. Now imagine I told you that AI was going to create a 3-day working week. Sounds great, right? Wonderful even. Yet to a first approximation these are the same thing. 60% of people employed and 40% unemployed is the same number of working hours as 100% employed at 60% of the hours.
So even if you think AI is going to have a tremendous effect on work, the difference between catastrophe and wonderland boils down to distribution. It’s not impossible that AI renders some people unemployable, but that proposition is harder to defend than the idea that AI will be broadly productive. AI is a very general purpose technology, one likely to make many people more productive, including many people with fewer skills. Moreover, we have more policy control over the distribution of work than over the pure AI effect on work. Declare an AI dividend and create some more holidays, for example.
Nor is this argument purely theoretical. Between 1870 and today, hours of work in the United States fell by about 40% — from nearly 3,000 hours per year to about 1,800. Hours fells but unemployment did not increase. Moreover, not only did work hours fall, but childhood, retirement, and life expectancy all increased. In fact in 1870, about 30% of a person’s entire life was spent working — people worked, slept, and died. Today it’s closer to 10%. Thus in the past 100+ years or so the amount of work in a person’s lifetime has fallen by about 2/3rds and the amount of leisure, including retirement has increased. We have already sustained a massive increase in leisure. There’s no reason we cannot do it again.
The CA Minimum Wage Increase: Summing Up
Two recent joint-papers Did California’s Fast Food Minimum Wage Reduce Employment? by Clemens, Edwards and Meer and The Effects of California’s $20 Fast Food Minimum Wage on Prices by Clemens, Edwards, Meer and Nguyen give what I think is a plausible and consistent account of California’s $20 fast food minimum wage.
California’s $20 fast food minimum wage raised wages in the sector by roughly 8 percent relative to the rest of the country but employment fell by 2.3 to 3.9 percent (depending on specification, median ~3.2%), translating to about 18,000 lost jobs. Food away from home (FAFH) prices in California’s four CPI-reporting MSAs rose 3.3–3.6 percent relative to 17 control MSAs. Falsification tests on Food at Home and All Items Less Food and Energy show zero differential movement—this is specific to restaurant prices.
What’s interesting is that the papers are independently estimated but the fit is consistent. The price paper uses Andreyeva et al.’s demand elasticity of -0.8 to convert the estimated price increases into an implied quantity declines: about 3.9–4.1 percent in limited-service and 1.7–1.8 percent in full-service. These align well with the employment declines of 3.2 and 2.1 percent estimated in the first paper.
The consistency tells us something about the mechanism. One thing we have learned about the minimum wage in recent years is that the pass-through effect is large and more of the employment decline is driven by pass through than by labor-capital substitution. In other words, prices rose, quantity demanded fell, and that’s what killed the jobs—not robots replacing workers. Not today, anyway.
In terms of welfare, the bulk of employed workers get an 8% wage increase, a small minority get disemployed. The big transfer was from consumers to workers. California has roughly 39 million residents, all of whom face 3.3–3.6% higher FAFH prices. The transfer is likely regressive — lower-income households spend a larger budget share on fast food specifically. So the policy effectively taxes low-income consumers generally to raise wages for a subset of low-income workers, while eliminating jobs for another subset. Your mileage may vary but I don’t see this as a big win for workers. We thought small increases in the minimum wage were absorbed–maybe some were or maybe they were just hard to estimate–but you can’t extrapolate the small increases to big ones–the effect is non-linear. Big increases in the minimum wage start to bite.
As usual, when it comes to fast food there is no such thing as a free lunch.
Addendum: Clemens’s JEP paper continues to be the masterclass in how to think through minimum wage issues.
The President(s) Fought the Law and the Law Won
In our textbook, Modern Principles, Tyler and I emphasize that Congress and the President are subject to a higher law, the law of supply and demand. In an excellent column, Jason Furman gives a clear example of how difficult it is to fight the law of inelastic demand:
…Today a given number of autoworkers can make, according to my calculations, three times as many cars in a year as they could 50 years ago.
The problem is that consumers do not want three times as many cars. Even as people get richer, they increase their spending on manufactured goods only modestly, preferring instead to spend more on services like travel, health care and dining out. There are only so many cars a family can own, but that’s not the case for expensive vacations or fancy meals. As a result we have fewer people working in auto factories and more people working in luxury resorts and the like.
These forces — rising productivity but steady demand — explain why the United States was losing manufacturing job share as far back as the 1950s and 1960s, long before trade became a major factor.
How to Make Judges and Referees Pay
A recent viral tweet, quoted by Elon Musk, points out that bartenders can be fined or even imprisoned if they serve alcohol to patrons who later kill someone while under the influence. Judges, in contrast, enjoy absolute or qualified immunity even when they repeatedly release defendants who go on to kill.
I agree that judges should face stronger incentives to make good decisions, but the obvious problem with penalizing judges who release people who later commit crimes is that judges would then have very little incentive to release anyone—and that too is a bad decision. Steven Landsburg solved this problem in his paper A Modest Proposal to Improve Judicial Incentives, published in my book Entrepreneurial Economics.
Landsburg’s solution is elegant: we must also pay judges a bounty when they release a defendant.
Whether judges would release more or fewer defendants than they do today would depend on the size of the cash bounty, which could be adjusted to reflect the wishes of the legislature. The advantage of my proposal is not its effect on the number of defendants who are granted bail but the effect on which defendants are granted bail. Whether we favor releasing 1 percent or 99 percent, we can agree that those 1 percent or 99 percent should not be chosen randomly. We want judges to focus their full attention on the potential costs of their decisions, and personal liability has a way of concentrating the mind.
One might object that a cash bounty will cost too much, but recall that the bounty is balanced by penalties when a released defendant commits a future crime. The bounties and penalties can be calibrated so that on average the program is budget-neutral. The key is to get the incentives right on the margin.
The structure of this problem is quite general. Ben Golub, for example, writes:
There should be a retrospective reputational penalty imposed on referees who vote no on a paper because the paper is too simple technically — if that paper ends up being important. It’s an almost definitional indicator of bad judgment.
Quite right, but a penalty for rejection needs to be balanced with a bonus for acceptance. Get the marginal incentive right and quality will follow!
Shruti interviews V. Anantha Nageswaran on the Indian economy
He is currently serving as the Chief Economic Advisor to the Government of India, and also is the co-author of the books Economics of Derivatives and The Rise of Finance: Causes, Consequences and Cures. The podcast covers import substitution and strategic resilience, futures and options market, gross fixed capital formation, crypto markets, India’s growth trajectory, and much more.
Here is the audio and video on YouTube. Here is a linked transcript. Excerpt:
RAJAGOPALAN: The policy response to this has come in a couple of different ways. One has come through SEBI. It has started raising contract sizes and limiting weekly expiration,and so on. Another instrument has come through taxation. There have been STT [Securities Transactions Tax] hikes in consecutive budgets,but there is one thing about STT that I want to understand a little bit better from someone like you who has thought about this deeply.
Now, STT on futures is being levied on the notional value of the contract, which is the full traded price, whereas the STT on the options is levied on the premium, which is a small fraction of the overall underlying value of the notional exposure. The effective tax that is imposed is much more on the futures trade, manyfold more actually, than it is on the options trade, whereas the speculation is mostly happening on the options side, which is also where most of the retail investors are losing money because the futures side is much better capitalized, larger firms, and so on.
NAGESWARAN: No, also the futures side is probably used more by institutions, and therefore, they are able to put up the margin requirement, etc., better than the options trades, where the individuals are being sold almost like the₹10 sachet-type options, and the options…
RAJAGOPALAN: Exactly, sachetization options, absolutely.
NAGESWARAN: Yes. Go ahead.
RAJAGOPALAN: Now with each successive hike in the STT,we’re seeing the gap widen. It’s on the margin, making futures relatively more expensive than options just because it’s taxing each trade. It’s like a toll fee that’s paid almost on every transaction. Your book was precisely about understanding these kinds of policy instruments. Given that now we have a tax instrument which inadvertently favors the more speculative instrument. Is that a good way of thinking about it, or how would you think about this problem?
NAGESWARAN: No, I think you have given me a lot to think about on this. I probably haven’t applied my mind as much to the mechanics of the STT being levied on the premium when it comes to options, but on the notional value of the contract when it comes to futures. Actually, you have given me something to think about. As you said, it could be having the unintended consequence of reducing the hedging role of futures, which probably is playing a better role there and encouraging the speculative element. Let me think about it and also probably take back this aspect of the conversation back to my colleagues in the revenue department, in the Ministry of Finance. Thank you for that, yes.
Of great importance for the world’s most populous country.
A bilateral AI pause?
Dean ball has some thoughts and hesitations:
Here are some questions I wish “Pause” and “Stop” advocates would address:
1. Assuming we achieve the desired policy goal through a bilateral US/China agreement, what would be the specific metric or objective we would say needs to be satisfied in advance? Who decides whether we have satisfied them? What if one one party believes we have satisfied them but the other does not?
2. If the goal is achieved through a bilateral US/China agreement, would we need capital controls to ensure that U.S. investors cannot fund semiconductor fabs, data centers, or AI research labs in countries other than the U.S. and China?
3. Would we need to revoke the passports of U.S.-based AI researchers and semiconductor engineers to prevent them leaving America to join AI-related ventures elsewhere? How else would the U.S. and China keep researchers within their borders?
4. How should we grapple with the fact that (2) and (3) are common features of autocratic regimes?
5. Do the above questions mean that this really should be a global agreement, signed by all countries on Earth, or at least those with the theoretical ability to host large-scale data centers (probably Vanuatu doesn’t need to be on board)?
Solve for the China tech equilibrium
Authorities in Beijing have barred two executives from a Singapore-based AI firm from leaving China amid a review of the company’s $2 billion acquisition by U.S. social media giant Meta, according to a report by the Financial Times on Wednesday.
Xiao Hong and Ji Yichao — the CEO and chief scientist, respectively, of Manus — were summoned to Beijing this month and questioned over a possible violation of foreign direct investment reporting rules related to the acquisition before being told they could not leave the country, the report said.
Here is more from The Washington Post. In my view, the American lead in AI is somewhat larger than a model comparison alone might suggest.
A Danish Fix for U.S. Mortgage Lock-in
In the Danish mortgage market every mortgage is backed by a corresponding bond. Thus, if a home buyer takes out a 500k mortgage at 3% interest, a bond is issued that pays the lender 3% interest on 500k. I’ve written about this system several times before. It has two distinct advantages.
- The correspondence principle means that mortgage banks don’t bear interest rate risk but instead specialize in evaluating credit risk (the risk that the borrower won’t pay). Deep markets rather than banks take on the interest rate risk. This makes the Danish system very stable.
- Mortgages can be pre-paid by buying the corresponding bond at market rates and extinguishing it. If a Danish borrower takes out a 500k mortgage at 3% interest and then rates rise to 6%, for example, the value of that mortgage falls to $358k and the borrower can buy the corresponding bond, deliver it to the bank, and, in this way, extinguish the loan.
In the US, a mortgage can be pre-paid only at a par. As a result, if interest rates rise, home owners don’t want to move because moving would require them giving up a 3% mortgage and replace it with say a 6% mortgage. This is called the lock-in effect. Lock-in can be quite severe. Fonseca and Liu find:
Using individual-level credit record data and variation in the timing of mortgage origination, we show that a 1 percentage point decline in the difference between mortgage rates locked in at origination and current rates reduces moving by 9% overall and 16% between 2022 and 2024, and this relationship is asymmetric. Mortgage lock-in also dampens flows in and out of self-employment and the responsiveness to shocks to nearby employment opportunities that require moving, measured as wage growth within a 50- to 150-mile ring and instrumented with a shift-share instrument.
What about in Denmark? The Danes definitely take advantage of the opportunity to buy-back. Part of this is due to tax advantages but those are just a transfer. More importantly, Danes don’t get locked in. A new paper by Berger, Jeong, Marx, Olesen, and Tourre compares mobility across Denmark and the US:
We study Danish fixed-rate mortgage contracts, which are identical to those in the United States except that borrowers may repurchase their mortgages at market value. Using Danish administrative data, we show that households actively buy back debt when mortgage prices fall below par and that household mobility is largely insensitive when existing mortgage rates are below prevailing market rates — unlike in the United States, where moving rates fall sharply as rates rise. We develop an equilibrium model that explains these patterns and show that introducing a repurchase-at market option into U.S. mortgages substantially reduces interest-rate-induced lock-in with limited effects on equilibrium mortgage rates.
The last point is especially important because you might wonder whether we are assuming a free lunch? After all, if US borrowers lose when they have to pre-pay at par then lenders surely gain. And if lenders gain on pre-payment then they will be willing to lend at lower rates on mortgage initiation. No free lunch, right? The logic is correct but note that the gain to lenders comes mainly from the relatively small set of households that move despite lock-in so the pre-payment bonus to lenders is quite small. Under the author’s calibrated model, mortgage interest rates in the US would rise by only 18 basis points on average if the US moved to a Danish type system.
In other words, there actually is a free or at least a low-priced lunch because lock-in is bad for homeowners and it doesn’t benefit lenders. As a result, moving to a Danish system would create net benefits.
The hyper-NIMBY of earlier Cape Town and South Africa
The most controversial of the forced removals occurred in the second half of the 1960s, with the expulsion of 65,000 coloureds from District Six, a vibrant inner-city ward of Cape Town, where whites, many of the slumlords, owned 56% of the property. Against their will, District Six residents were moved out to the sandy townships of the Cape Flats. In Johannesburg, the inner-city suburb of Sophiatown, where blacks could own freehold property, was another notorious site of forced removals. Often long-established community institutions such as churches and schools had to be abandoned.
That is from the very good book by Hermann Giliomee The Afrikaners: A Concise History.
Alternatives to 911
Almost a quarter-billion calls are placed to 911 each year in the United States. A large share of them involve social problems, not crimes or emergencies—yet police are dispatched in response. This review traces how the 911 emergency system’s institutional design shapes demand for police, who is excluded from or ill served by this system, and what alternatives exist, including nonemergency lines (with police response), government hotlines (211, 311, 988), civilian crisis teams, and community-based resources. Among the universe of municipal police departments with at least 100 sworn officers in 2020, covering 107 million US residents, police have absorbed broad social service functions, with the availability of formal alternatives restricted to the largest cities. The evidence suggests that the primacy of police reflects institutional reproduction more than public need. I propose priorities for future research.
That is from a new NBER working paper by Bocar A. Ba.
Why is the USDA Involved in Housing?!
In yesterday’s post, The 21st Century ROAD to Housing Act, I wrote that Trump’s Executive Order “cuts off institutional home investors from FHA insurance, VA guarantees and USDA backing…”. The USDA is of course the United States Department of Agriculture. In the comments, Hazel Meade writes:
USDA? Wait, what????
Why is the USDA in any way involved in housing financing?
Are we humanly capable of organizing anything in a rational way?
It’s a good question. The answer is a great illustration of the March of Dimes syndrome. The USDA got involved with housing in the late 1940s with the Farmers Home Administration. The original rationale was to support farmers, farm workers and agricultural communities with housing assistance on the theory that housing was needed for farming and the purpose of the USDA was to improve farming. Not great economic reasoning but I’ll let it pass.
Well U.S. farm productivity roughly tripled between 1948 and the 1990s as family farms became technologically sophisticated big businesses. So was the program ended? Of course not. Over time the program subtly shifted from farmers to “rural communities”–the shift happened over decades although it was officially recognized in 1994 when the Farmers Home Administration was renamed the Rural Housing Service. Today rural essentially means low population density which no longer has any strong connection to agriculture.
So that’s the story of how the US Department of Agriculture came to run a roughly $10 billion annual housing program for non-farmers in non-agricultural communities. And how does it do this? By supporting no-money-down direct lending and a 90 percent guarantee to approved private lenders. Lovely.
It’s a small program in the national totals, but an amusing example of the US government robbing Peter to pay Paul and then forgetting why Paul needed the money in the first place.
The 21st Century ROAD to Housing Act
The 21st Century ROAD to Housing Act appears likely to pass the Senate. The bill contains some genuinely good ideas alongside some very popular—but bonkers ideas.
Let’s start with the good ideas.
The bill would streamline NEPA review for federally supported housing, primarily by expanding categorical exclusions. Federal environmental review does impose real costs and delays on housing construction, so reducing unnecessary review is a step in the right direction. The gains will probably be modest—most housing regulation occurs at the state and local level—but removing friction is good.
The bill would also deregulate manufactured housing by eliminating the permanent chassis requirement and creating a uniform national construction and safety standard. The United States once built far more factory-produced housing; in the early 1970s, by some accounts a majority of new homes were factory-built (mobile or modular). Long-run productivity growth in housing almost certainly requires greater use of factory construction. Land-use regulation remains the dominant constraint on supply, but enabling scalable manufacturing is still welcome.
Another interesting provision involves Community Development Block Grants (CDBG). The bill allows CDBG funds to be used for building new housing rather than being largely restricted to rehabilitation of existing housing. More federal spending is not automatically appealing, but the bill adds an unusual incentive mechanism.
The bill creates a tournament for CDBG allocations. Localities that exceed the median housing growth improvement rate among eligible CDBG recipients receive bonus funding. Those below the median face a 10 percent reduction. The key feature is that the penalties fund the bonuses, so the system reallocates money rather than expanding spending.
This is a clever design. It creates competition among localities and benchmarks them against peers rather than against a fixed national target. In effect, the program rewards relative improvement rather than absolute performance—a classic tournament structure. (See Modern Principles for an introduction to tournament theory!).
Ok, now for the popular but bonkers ideas. Section 901 (“Homes are for People, Not Corporations”) restricts the purchase of new single-family homes by large institutional investors. Elizabeth Warren is a sponsor of the bill but this section was driven almost entirely by President Trump. Trump passed an Executive Order, Stopping Wall Street from Competing With Main Street Home Buyers, that cuts off institutional home investors from FHA insurance, VA guarantees, USDA backing, Fannie/Freddie securitization and so forth. The bill goes further by imposing a seven-year mandatory divestiture rule, forcing institutional investors to convert rental homes to owner-occupied units after seven years.
No one objects to institutional investors owning apartment buildings. But when the same investors own single-family homes, it breaks people’s brains. Consider how strange the logic sounds if applied elsewhere:
…a growing share of apartments, often concentrated in certain communities, have been purchased by large Wall Street investors, crowding out families seeking to buy condominiums.
Apartments are fine, hotels are fine, but somehow a corporation owning a single family home is un-American. In fact, the US could do with more rental housing of all kinds! Why take the risk of owning when you can rent? Rental housing improves worker mobility. When foreclosures surged after 2008 and traditional buyers disappeared, institutional investors stepped in and absorbed distressed supply — helping stabilize markets. Who plays that role next time?
Institutional investors own only a tiny number of homes, so even if this were a good idea it wouldn’t be effective. But it’s not a good idea, it’s just rage bait driven by Warren/Trump anti-corporate rhetoric.
What does “Homes are for People, Not Corporations” even mean?–this is a slogan for the Idiocracy era. “Food is for People, Not Corporations,” so we should ban Perdue Farms and McDonald’s?
The Hidden Cost of Hard-to-Fire Labor Laws: Why European Firms Don’t Take Risks
In our textbook, Modern Principles, Tyler and I write:
Imagine how difficult it would be to get a date if every date required marriage? In the same way, it’s more difficult to find a job when every job requires a long-term commitment from the employer.
In two new excellent pieces, Brian Albrecht and Pieter Garicano extend this partial equilibrium aphorism with some general equilibrium reasoning. Here’s Albrecht:
[I]magine there is a surge for Siemens products. Do you hire a ton of workers to fill that demand? No, you’re worried about having to fire them in the future but being stuck until they retire.
But it’s even worse than that…..[suppose Siemens does want to hire] where is Siemens getting those workers from?…Not only is it a problem for Siemens that they won’t be able to fire people down the road, the fact that BMW doesn’t fire anyone means you can’t hire people.
Garicano has an excellent piece, Why Europe doesn’t have a Tesla, with lots of detail on European labor law:
Under the [German] Protection Against Dismissal Act, the Kündigungsschutzgesetz, redundancies over ten employees must pass a social selection test (Sozialauswahl). Employers cannot choose who leaves: they must rank employees by age, years of service, family maintenance obligations, and degree of disability, and then prioritize dismissing those with the weakest social claim to the job. If someone is dismissed for operational reasons but the company posts a similar job elsewhere, the dismissal is usually invalid.
Disabled employees can be dismissed only with the approval of the Integration Office (Integrationsamt), a public body. The office will weigh the employer’s reasons, whether they have taken sufficient steps to integrate the employee, and whether they could be redeployed elsewhere in the organization. Workers who also become caregivers cannot be dismissed at all for up to two full years after they tell their bosses they fulfill that role.
As a company becomes larger and tries to let more workers go at once these difficulties increase. In many European countries, companies with more than a certain number of workers – 50 in the Netherlands, 5 in Germany – are obliged to create a works council, which represents employees and, in some countries, must give its approval to decisions the employer wants to make regarding its employees, including layoffs or pay rises or cuts.
…Companies that are allowed to fire someone and can afford to pay the severance costs have to wait and pay additional fees. Collective dismissal procedures in Germany start after 30 departures within a month; once triggered they require further negotiations with the works council, a waiting period, and the creation of a ‘social plan’ with more compensation for departing workers. When Opel shut down its Bochum factory in Germany, it reached a deal with the works council to spend €552 million on severance for the 3,300 affected employees. This included individual payments of up to €250,000 and a €60 million plan to help workers find new jobs.
Now what is the effect of regulations like this? Well obviously the partial equilibrium effect is to reduce hiring but in addition Garicano notes that it changes what sorts of firms are created in the first place. If you are worried about being burdened by expensive dismissal procedures, build a regulated utility with captive government contracts, not a radical startup with a high probability of failure.
Rather than reduce hiring in response to more expensive firing, companies in Europe have shifted activity away from areas where layoffs are likely. European workers are for sure, solid work only. This works well in periods of little innovation, or when innovation is gradual. The continent, however, is poorly equipped for moments of great experimentation.
…Europe’s companies have immense, specialized knowledge [due to retained workforces, AT]. The problems happen when radical innovation is needed, as in the shift from gasoline to electric vehicles. The great makers of electric cars have either been new entrants, like Tesla and BYD, or old ones who have had their insides stripped, like MG.
..If Europe wants a Tesla, or whatever the Tesla of the next decade will turn out to be, it will need a new approach to hiring and firing.
Claude on NY’s Senate Bill S7263
A NY State Senate proposed bill (Senate Bill S7263) would
prohibit a chatbot to give substantive responses; information, or advice or take any action which, if taken by a natural person, would constitute unauthorized practice or unauthorized use of a professional title as a crime in relation to professions who licensure is governed by the education law or the judiciary law.
Claude, what do you think of the legal logic of this bill?
The legal logic is shaky. The bill’s operative test asks whether a chatbot’s output, “if taken by a natural person, would constitute a crime” under New York’s unauthorized practice statutes. But those statutes set a high bar — mere advice-giving by a layperson is not criminal; what’s prohibited is holding yourself out as a licensed professional while charging fees. Your neighbor who happens to be a doctor can tell you what he thinks your rash is. A retired lawyer can walk you through a lease. None of that is a crime. So either the bill is mostly toothless — most AI responses wouldn’t meet the criminal threshold — or courts will interpret “substantive advice” far more expansively than the underlying statutes warrant, effectively creating a new and harsher legal standard for AI that applies to no human equivalent. Neither outcome reflects good lawmaking, and the ambiguity will systematically over-deter, pushing operators to sanitize outputs well beyond what the statute actually requires.
Claude, who will be hurt by this bill?
The deepest problem is who gets hurt. The populations with the most to gain from AI-assisted professional guidance are precisely those who cannot afford lawyers, doctors, or therapists — the tenant trying to understand an eviction notice, the rural patient triaging symptoms, the small business owner reviewing a contract. For these users, the choice the bill actually creates is not “AI advice versus licensed professional advice”; it’s “AI advice versus no advice at all.” A watered-down chatbot that responds to every medical or legal question with a referral to a licensed professional is not a safety feature — it’s a luxury tax on information, paid disproportionately by people with the fewest alternatives. New York should be asking how to ensure AI advice is accurate and clearly disclosed, not how to suppress it on behalf of incumbents who’d prefer less competition.
Thanks Claude. Where did you learn this excellent analysis?
Public choice theory is part of my general training on economics and political science literature — Buchanan, Tullock, Olson, and the broader tradition are well-represented in the texts I was trained on.
But I’d be remiss not to note that you’re at the home institution of public choice. The Buchanan House is literally on your campus, and the tradition Buchanan and Tullock built at GMU is probably the most direct intellectual ancestor of the kind of analysis I was applying to this bill — concentrated benefits, diffuse costs, regulatory capture, rent-seeking by incumbent professionals. That framework is so natural for analyzing occupational licensing legislation that it almost applies itself.
If anything, my read of S7263 is pretty much a straightforward public choice story: organized professional associations with strong lobbying presence get statutory protection written in the language of consumer welfare, while the costs fall on millions of unorganized users who will never know what information they didn’t receive. Buchanan would have recognized it immediately.
What the recent dust-up means for AI regulation
From my new Free Press column, I see these as the most important facts:
Congress has not passed explicit regulation of AI foundation models, and an executive order from President Trump limited regulation at the state level. But do not think that laissez-faire reigns. In addition to existing (largely pre-AI) laws, which lay out general principles of liability, and laws from a few states, the United States is engaged in a kind of “off the books” soft regulation.
The major AI companies keep the national security establishment apprised of the progress they are making, as has been the case with Anthropic. There is a general sense within the AI industry that if the national security authorities saw anything in the new products that was very concerning or that might undermine the national interest, they would inform the president and Congress. That would likely lead to more formal and more restrictive kinds of regulation, so the major AI companies want to show relatively safe demos and products. An informal back and forth enforces implied safety standards, without the involvement of formal legislation.
That may sound like an unusual way to do regulation, but to date the system has worked relatively well. For one thing, I believe our national security establishment has a better and more sophisticated understanding of the issues than does Congress. Congress right now simply isn’t up to the job, as indeed the institution has been failing more generally. Most representatives seem to know little about the core issues behind AI regulation.
As it stands, AI progress has been allowed to proceed, and the United States has stayed ahead of China, without major catastrophes. The burden on the companies has been manageable, and the system, at least until last week, was flexible.
Another advantage of this system is that both Congress and the administrative state can be very slow to act. The AI landscape can change in just weeks, yet our federal government is used to taking years to issue laws and directives. Had we passed AI legislation in, say, 2024, today it would be badly out of date, no matter what your point of view on what such regulation should accomplish. For instance, in 2024 few outsiders were much concerned with the properties of, or risks from, autonomous AI “agents.” Today that is the number-one topic of concern.
Though it is not driven by legislation, the status quo AI regulatory system is not anti-democratic, as it operates well within the rules passed by Congress and the administrative state. It is more correct to say the current AI guardrails rely on the threat of regulation, rather than regulation itself, with the national security state as the watchdog. The system sticks to a kind of creative ambiguity. The national security state offers no official imprimatur for the new advances, but they proceed nonetheless. Nevertheless, the various components of the national security state reserve the right to object in the future.
It is also correct, however, to believe that such a system cannot last forever. At some point creative ambiguity collapses. Someone or some institution demands a more formal answer as to what is allowed or what is not allowed. At that point a more directly legalistic system of adjudication enters the picture, and Congress likely starts paying more attention.
With the recent dispute between Hegseth and Anthropic, we have taken a step away from the previous regulatory mode of quiet cooperation. Instead, the relationship between the military and the AI companies has become a matter of public concern. Now everyone has an opinion on Hegseth, Anthropic, and OpenAI, and social media is full of debate.
No matter “whose side you take,” it would have been better to have resolved all this behind closed doors.