After decades of effort, the federal government has finally gotten the message on geothermal power. Over the past several years, it has:

1. Authorized and funded EGS pilot demonstrations,

2. Extended geothermal investment and production tax credits,

3. Approved the 2 GW Fervo Cape project and held the largest BLM geothermal lease sale in over 15 years,

4. Established categorical exclusions under NEPA to streamline geothermal permitting by up to a year,

5. Made billions in potential loan authority available through the Office of Energy Dominance Financing,

6. Funded early-stage drilling and closed-loop innovation through ARPA-E,

7. Launched the FedGeo technical assistance initiative and Geothermal Power Accelerator with NASEO,

8. Elevated geothermal to a named priority in the energy emergency declaration,

9. And restructured DOE to house geothermal alongside hydrocarbons.

The combination of these efforts has led to the most favorable regulatory environment in decades. And, unlike the rest of the energy industry, which has reacted to the current demand spike with horror, geothermal is taking full advantage.

Given this fortuitous environment, I thought it would be worth taking the time to examine what state issues are most likely to block this rollout in an attempt to head them off at the pass. This post will focus exclusively on state-level burdens, as federal burdens have (1) already been significantly reduced, and (2) are already well covered by other policy work. It will also not cover areas that, while important to geothermal (e.g. interconnection and permitting reform), are broader reforms that just so happen to help geothermal.

To put it simply, geothermal’s problem is that state laws and regulations were written for industries that extract things from the ground—mostly oil, gas, water, and minerals. New geothermal projects, from closed-loop to EGS, do not share this extractive nature. This makes them a poor fit for such laws and regulations. Because the fit to any particular category is poor, states classify geothermal resources haphazardly, open geothermal projects up to water rights adjudication risk, subject geothermal projects to onerous standards that don’t match their environmental footprint, and leave them in limbo about which agencies are responsible for permitting them.

None of these are the fault of any state in particular. The nature of any new technology is to share certain attributes with the old and differ on others. However, in the case of geothermal, states are currently choosing so differently that the patchwork currently being quilted is likely to serve as a serious barrier to scaling. In an attempt to provide clarity, the following will walk through the problems listed above step-by-step, starting with classification, discuss the potential solutions on offer, and then provide a potential solution for each. Solutions should be read as suggestions rather than firm reform recommendations here.

Subscribe now

Classification

Before a geothermal project can be permitted, the state has to decide what the resource is. In most Western states, this determination is usually between one of two separate categories: surface estates and mineral estates. For a given piece of property, the surface estate covers the land itself; the mineral estate covers subsurface resources like oil, gas, and coal. The issue is that these estates are routinely held by different people. A rancher may own the surface while a mining company or distant heir owns everything below it. Geothermal does not fit cleanly into either estate, which is why states have reached for different analogies and ended up in different places.

Mineral estate states. California classifies geothermal as a mineral right, which means a developer must secure the mineral owner’s consent even if the surface owner wants the project. In split-estate parcels, that mineral owner may have acquired rights for oil or mining purposes and have no interest in geothermal, or may be hard to locate at all. Title searches, the process by which an individual determines who owns the property in question, are already expensive for oil and gas; for geothermal, where the legal basis for treating heat as a “mineral” is less settled, they carry the added risk that a court later disagrees with the classification as conveyed in a particular deed.

Surface estate states. Nevada and Washington take the opposite approach, giving geothermal rights to the surface landowner. This simplifies ownership, as locating surface owners is a much easier process. However, a mineral owner may still argue that geothermal drilling or fluid circulation at depth interferes with their subsurface rights.

Hybrid approaches. Other states have tried to split the difference. Colorado divides jurisdiction by depth: shallow systems stay under the State Engineer, while anything below 2,500 feet goes to the Energy and Carbon Management Commission. Texas consolidated everything under the Railroad Commission. Wyoming treats geothermal as underground water. Idaho, Montana, and Virginia each created standalone definitions that do not map onto any of these frameworks.

As you may have surmised, this system is a mess. Each classification produces a different set of parties who must consent before a project proceeds, and a different agency door to walk through. Under mineral-estate rules, developers negotiate with mineral holders and their lessees. Under surface-estate rules, the parties are easier to identify but subsurface conflicts persist. Under hybrid rules, a developer may need legal analysis just to determine which regime applies. When the industry was incubating, this was not an issue, as the need to scale across state lines was minimal. However, as it scales, this level of diversity will cause a substantial drag.

Potential Solution. One potential fix here is a voluntary model code, analogous to the UCC (Uniform Commercial Code), developed through the NASEO Accelerator with participating states. It should address at minimum whether geothermal attaches to the surface or mineral estate, how split-estate conflicts are resolved, and how the classification interacts with water law. A model code will give state legislatures a common reference point to adopt rather than forcing each to invent from scratch and going in completely different directions.

Water Rights

There is an additional wrinkle with the classification problem from section (1). There is an area of the law called water law, and every major Western geothermal state operates under a doctrine particular to it called prior appropriation. That doctrine says that the first user to put a flow of water to use holds the priority claim, and no latecomer can infringe upon their supply. This system was intended to manage the limited water supply in water-parched Western states. In theory, this shouldn’t matter for new geothermal systems, as reinjection withdraws fluid, extracts heat, and returns the fluid to the same formation. It does not consume the water in any meaningful quantity (except up front when creating the fractures), but prior appropriation does not cleanly distinguish between taking water out and passing it through.

Unfortunately, this gap means that developers who do not have a specific exception to point to face substantial legal risk. For a technology whose economics depend on front-loaded drilling capital and long payback periods, that timeline can make the project unfinanceable. Lenders are unlikely to underwrite a schedule that includes the possibility of a multi-year water rights adjudication.

States have tried to address this issue, but the solutions have been unstable so far. Nevada exempts reinjection-based operations from water appropriation permits on the theory that non-consumptive use does not constitute an appropriation. Assembly Bill 109 last year (2025) attempted to revoke that exemption, but did not advance. California’s Sustainable Groundwater Management Act may restrict geothermal activity where local agencies determine that sustainable yield is threatened. New Mexico explicitly authorizes existing water-rights holders to bring impairment actions against geothermal developers.

Potential Solution. States issue categorical determinations that reinjection-based geothermal operations are non-consumptive, paired with a statutory safe harbor preventing retroactive reversal. Prior appropriation already recognizes non-consumptive uses in other contexts, so the doctrinal basis exists. The safe harbor is the critical piece. Without it, developers face the Nevada problem, where one legislature’s classification can potentially be undone by the next, a real risk for such a capital intensive endeavor.

*There is a caveat here. Some geothermal operations do affect aquifer levels and temperatures even with reinjection, so any exemption should include monitoring tied to measurable aquifer impacts, with a mechanism for case-by-case review and liability imposition where warranted.

Permitting

The classification and water rights problems feed directly into a third. It’s everyone’s favorite issue: permitting. Given that most states have no geothermal-specific permitting process, the permits required for geothermal projects were almost certainly designed for the category it was placed in, not for geothermal itself. As a result, the forms, bonding requirements, review timelines, and compliance obligations reflect the risks of the inherited category rather than the risks geothermal actually presents. That mismatch then compounds across the multiple phases of a geothermal project, each of which imposes obligations intended to compensate for a different technology. The combination of these features makes geothermal permitting dramatically more burdensome than it should be.

Exploration. Exploration is where the gap between bureaucratic requirements and real environmental impact is largest. Site assessment (e.g. temperature gradient holes, seismic surveys, geologic mapping) is exceedingly light work, so light that much of it can be conducted from the back of an ATV. But in states regulating geothermal under oil and gas or mining law, a temperature gradient hole drilled to measure heat flow triggers the same application, bonding, and review as a production well. Those requirements are an enormous cost burden on a process that often needs to be run dozens of times to find a good drill-hole.

Drilling. This is the one phase where borrowing is largely appropriate. Deep drilling involves high pressures and wellbore integrity risks that are similar whether the well targets oil or heat. Small tweaks may need to be made to address additional heat and corrosion risks.

Operations. However, once the holes are drilled, a geothermal project diverges completely from a traditional drilling project. There is no commodity to extract, minimal produced water to dispose of, nor anything like the same spill risk. But for the most part, reporting, bonding, and compliance still assume that there is. Oil and gas forms are the most common example (forms sometimes even ask about projected hydrocarbon production), but the pattern repeats across every inherited category: consumptive-use appropriation permits for a non-consumptive activity, waste-disposal injection requirements for a sealed loop.

New York has kindly demonstrated both the problem and the difficulty of fixing it. Until 2023, closed-loop boreholes deeper than 500 feet required oil-and-gas drilling permits. Developers in the state were required to post well plugging bonds designed for oil and gas operations for sealed plastic pipes filled with antifreeze. Governor Hochul signed S. 6604/A. 6949 to exempt them, but the replacement geothermal-specific rules, due by December 2024, are still not in place as of this writing.

Potential Solution. States should draft flexible frameworks to tier permitting obligations to the actual risk at each phase. For example, exploration could require only notification and basic land-restoration commitments. Closed-loop boreholes could get a permit-by-rule structure. To implement such calibration, legislatures should designate a specific agency to develop geothermal-specific forms, bonding, and timelines within a statutory deadline with some teeth.

Agency Jurisdiction

Even if a developer knows what framework applies, there is often a much more basic problem: it is deeply unclear which agency is responsible for granting the permits. Because geothermal sits across several existing categories, it frequently lands in gaps between regulators, or in the overlap where several agencies each claim a piece and there is no single agency who owns the whole.

California, unsurprisingly, is the most instructive example. CalGEM (the state’s oil and gas regulator) permits geothermal resource wells on state and private land. The Department of Water Resources sets construction standards for heat exchange wells under a separate authority, which just so happen to be enforced by county departments. Then, local jurisdictions pile building, grading, and other well permits on top. Which agency controls a given project depends on well depth, operating temperature, and whether the system taps a defined geothermal reservoir or just ambient earth heat. That boundary is more than blurry enough to force a developer to approach three or four agencies before learning which one will actually process the application. And sometimes they will disagree on which agency this is.

Potential Solution. Statutorily pick one department to host all geothermal permitting.

Conclusion

The common thread through all four of these issues is straightforward: geothermal does not fit the categories states built, and nobody has built the right ones yet. This is good news! Political problems are tricky, policy ones less so. All we need is a little elbow grease. Time to get cracking.

Where New York Is Now

Permitting reform is finally on the table in New York. In her State of the State speech, Governor Hochul proposed the most significant reform to New York’s environmental review law (SEQRA) since it was enacted in 1975. The proposal would exempt housing projects of up to 250 units in New York City (500 in high-density areas), and up to 100 units elsewhere, from SEQRA review entirely.1 It would also impose statutory timelines on the environmental review process as a whole for the first time: one year to determine whether a full environmental impact statement is needed, two years to complete one. Given the scale of the proposal, it seems like a good time to give a brief explainer on what SEQRA is, what it requires, its structural flaws, and whether the proposed reforms are likely to succeed.

To start, the basics. SEQRA requires every state and local agency in New York to assess the environmental consequences of its actions before approving them. SEQRA borrowed its statutory structure from the National Environmental Policy Act of 1969, but with three major departures. It applies to local agencies, not just state ones. As a result, the hundreds of municipal planning boards and other public authorities that dot the state are all subject to environmental review obligations. It adopts a lower triggering threshold: review is required when an action “may have” a significant adverse impact, a probabilistic standard that forces agencies to act on possibility rather than certainty.2 And it embeds a substantive mandate that NEPA lacks. ECL section 8-0109 requires agencies to certify that adverse environmental effects will be “minimized or avoided to the maximum extent practicable.” This is the largest conceptual departure from NEPA, which federal courts read as purely procedural. New York’s statute, at least on its face, demands more.

Then there are the implementing regulations. These are codified at 6 NYCRR Part 617, and detail an in-depth sequential process of review. First, every action requiring agency discretion must be classified as Type I (presumptively significant, requiring a full environmental assessment form and coordinated review), Type II (categorically exempt), or Unlisted (everything else, requiring at minimum a short assessment and a significance determination). If the lead agency determines that the action may have a significant adverse impact, it issues what is called a “positive declaration,” triggering preparation of a draft environmental impact statement, public comment, a final EIS, and a findings statement certifying that impacts have been minimized. If it determines the action will not have significant impacts, it issues a negative declaration, and the project proceeds.3

The key problem with SEQRA is that this process is incredibly poorly matched with the structure of judicial review. The way SEQRA is set up, it both grants the agencies an enormous amount of discretion and makes the exercise of that discretion reviewable by a court. Each stage requires the lead agency to make a decision under an open-ended standard; however, each of these decisions is independently reviewable through an Article 78 proceeding in state court (at least if the outcome is a decision by the agency that the project can proceed). Further, each review is governed by the “hard look” doctrine established by the Court of Appeals in 1986.4 Under this standard, the agency must identify the relevant areas of environmental concern, take a hard look at them, and provide reasoned elaboration of its determination.5 In theory, this should be quite a deferential standard. The core question is “Did the agency reasonably use its discretion in the allocation of its resources for this specific project given its scope and scale?” However, in practice, these cases require case-specific examination of the administrative record. Such examination resists summary disposition, which means every challenge takes months to years to resolve regardless of its merit.

Subscribe now

SEQRA’s Contestable Area

In an attempt to compress this concept into a usable format, I have taken to referring to it as “contestable area”: the aggregate surface area of litigable discretionary decisions, governed by fact-intensive review standards, arrayed across sequential stages, each independently actionable. The concept is distinct from ordinary statutory ambiguity. Almost all laws are ambiguous around the edges of their application. What makes SEQRA’s (as well as other environmental statutes) ambiguity distinctive is the multiplicative interaction of sequentiality, cost asymmetry, and a review standard that resists quick resolution. In short, the space of attack for any SEQRA decision is so large that you can almost guarantee that there is a colorable lawsuit available to delay the agency decision.

To demonstrate this concept, it is useful to walk through an example. For example, let’s say a 90-unit mixed-use residential development is proposed on a previously commercial site in the lower Hudson Valley. There are two involved agencies: the town planning board and the county health department. The project starts off Unlisted, falling below the presumptive Type I residential threshold of 200 units. The neighbors are opposed, and the site has no extraordinary environmental sensitivity.

The first litigable decision is classification. In this case, the developer owns an adjacent parcel. If the “whole action” includes a potential second phase on that parcel, the combined unit count might exceed Type I thresholds, presumptively requiring an EIS. Noticing this, a challenger argues the board impermissibly segmented the project under 6 NYCRR section 617.3(g). The board says that the parcels are independent. The key factor here is whether the segmented piece has “no independent utility.” This is a fact-intensive determination, requiring the court to examine the record.

The second is lead agency designation. This one can be ignored, but it is worth noting as it consumes 30 to 60 days of calendar time.

The third is the environmental assessment form. The planning board rates traffic impacts “moderate,” stormwater “moderate,” neighborhood character “moderate.” In part 3 of the form, the section where the board must explain why these moderate impacts do not rise to significance, the board writes five pages. A challenger argues this is conclusory, not reasoned elaboration. They argue that the board failed to conduct a traffic study or consider cumulative impacts from three other projects approved within a mile. Each impact category is independently contestable.

The fourth is the significance determination itself. After its deliberations, the board issues a negative declaration. This is the largest contestable area. After Chinese Staff confirmed that “environment” included socioeconomic impacts such as displacement and neighborhood character, virtually any development project presents arguable grounds for challenge. The standard, “may have a significant adverse impact,” does not specify a probability threshold, nor has an appellate court supplied one. The result is that the standard accommodates nearly any claim of potential harm that a competent attorney can put on paper.

If the board decides to play it safe and proceed with a full EIS via issuing a positive declaration, there would be another crop of litigable decisions: whether the scope adequately captured all significant impact categories; whether the draft EIS analysis was sufficient; whether the final EIS meaningfully responded to public comments; and whether the findings statement’s certification that impacts were “minimized or avoided…to the maximum extent practicable” was supported by the record. The Court of Appeals held in Jackson v. New York State Urban Dev. Corp., 67 N.Y.2d 400 (1986), that this language imposes a genuine substantive obligation. This means that agencies must actually minimize harm, not merely document that they considered it. But the court simultaneously held that it would not “weigh the desirability of any action or choose among alternatives.” To be blunt, these two conclusions are inconsistent and their tension has never been clearly resolved.

The combination of all of these decision points creates an incredibly large contestable area. For the 90-unit project, six theories of insufficiency could in theory be put forward: on classification, on the lead agency determination, on the extent of the EAF analysis, on the significance determination, and on the two components of the underlying land-use approval. Each of these decisions may contain enough fact-specific use of discretion to generate a claim sufficient to survive early dismissal motions.

How SEQRA Interacts with Litigation

The combination is crucial because of two features of the litigation environment. The first is cost asymmetry. The plaintiff needs to merely plead insufficiency. The defendant, the agency, has to compile and certify the entire administrative record (which can run into the thousands of pages). While this is going on, the project developer must pay the carrying costs of a frozen project for the duration of the litigation. This asymmetry means the expected value of filing a challenge is positive for opponents even at low probabilities of success, as long as the delay value exceeds the filing cost. If you are a neighbor who would like to see a project die, it usually does. The second is the financing freeze that comes along with lawsuits. It is true that there is no automatic stay of agency action upon filing an Article 78 petition; the approval remains legally effective unless the court enjoins the decision.6 However, even if an injunction is not issued, good luck finding a construction lender that will close on a loan while active litigation threatens to annul the underlying approval. This effectively limits the ability to proceed during a case to developers who can self-finance while bearing extraordinary downside risk.7

Returning to the 90-unit project, let’s play out the sequence. The neighbors file an Article 78 challenging the negative declaration. Then, the lender suspends the loan commitment and the project is halted. Fourteen months later, the court partially sustains the challenge: the board failed to address cumulative traffic impacts from nearby projects. The board commissions a $40,000 traffic study, completes a supplemental analysis, and then issues a new negative declaration. The delays total twenty months. Say the cost to the developers is approximately $100,000 in legal fees plus carrying costs. To be fair to the neighbors, according to the court, the board’s initial review was genuinely deficient. But the fix was a single $40,000 study. The remaining $60,000 and fourteen months bought nothing except the privilege of proving that in court.

For some useful numbers, the Citizens Budget Commission found in a 2022 study of 171 private zoning applications filed in New York City between 2014 and 2017 that the median approval time was 2.5 years, with 80 percent of elapsed time consumed by pre-certification and environmental review. They estimated that such review added 11 to 16 percent to total project costs, which would be approximately $58,000 per unit in low-rise multifamily and $67,000 per unit in high-rise construction.

Reform and Hochul’s Proposal

So, returning to Hochul’s proposal, is it likely to succeed? In a word, yes. The two categories of housing mentioned in the proposal would be statutorily excluded from SEQRA review (functionally similar to being added to the Type II list). No classification, no EAF, no significance determination, no Article 78 SEQRA challenge. For the 90-unit project, the delay would be eliminated entirely. By eliminating the agencies’ discretion to say whether a project may have an environmental impact, the proposal eliminates its contestability.

However, the proposal is a blunt tool that does very little for non-qualifying projects. A 600-unit development, or a 200-unit project on a greenfield site, faces the same steps with the same open-ended standards as it did before. Mandatory timeline caps may constrain the duration of the process (though I have my doubts) but do not narrow the grounds for challenge or accelerate judicial resolution. If Governor Hochul wishes to truly reform SEQRA, it will require massively shrinking the contestable area the statute provides. A good general principle to hold is: “If courts are not going to reasonably defer to agency discretion, agencies should not be granted that discretion.”

Fixing SEQRA more broadly requires either shrinking the ability of courts to challenge agency discretion or shrinking the scale of that discretion. Either something like Article 78 pleading requirements for SEQRA need to be heightened (there are other legal remedies here, this is simply illustrative) or agencies need to be given a more ministerial role (which functionally is what Hochul’s proposal does for the categories it covers). There are methods for both, such as exhaustion requirements or objective statutory thresholds for positive and negative declarations, but whatever the method, the goal is clear: shrink the contestable area.

One of the well-commented-on oddities of environmental law is that the National Environmental Policy Act, the most litigated environmental statute in America, contains no provision for private litigation. A brief overview confirms this. Section 101 declares a national policy of preservation. Section 102 of the act directs federal agencies to prepare detailed environmental impact statements. But nowhere does the statute mention courts, private attorneys general, or judicial review. Nor is there any indication that the members of Congress who passed NEPA understood it to create private standing. When President Nixon signed the act on January 1, 1970, there was no doctrine by which a conservation group would have standing to haul an agency into federal court for failing to study the environmental consequences of its decisions.

And yet—within months—such suits became routine. More importantly, such suits were almost instantaneously afforded deference by the courts. This raises the question: How? How did doctrine change so suddenly to allow for previously uncognizable (e.g. environmental) injuries to provide not only standing but injunctions against government action?

The conventional answer is the Administrative Procedure Act, which permits persons “adversely affected or aggrieved” by agency action to seek judicial review. But the APA was passed in 1946, and for 24 years, environmental plaintiffs were not seen to be “aggrieved” by agency action. Before 1970, standing doctrine required what is referred to as a “legal interest”—a right that was recognized at common law or in statute. Back then, if an organization wanted to stop a ski resort being built to protect the natural beauty of the area, it faced a problem: no law gave it the right to enjoy said beauty. If you wanted to challenge such an act, you needed to win an election.

However, this standard was transformed in March 1970, when Justice William O. Douglas announced in Association of Data Processing Service Organizations v. Camp that standing required not a legal interest but an “injury in fact, economic or otherwise.” That case, aside from being a mess that the Court would spend decades cabining, was the doctrinal key that unlocked NEPA litigation. It turned out that fitting an injury underneath the category “otherwise” was not overly difficult.

Subscribe now

But where did Douglas find this formula? In keeping with the rest of his jurisprudence, the opinion offers little by way of explanation. However, in his opinion, he did cite two circuit court decisions from the mid-1960s for the proposition that standing could rest on “aesthetic, conservational, and recreational” values. The cases were Scenic Hudson Preservation Conference v. Federal Power Commission, decided by the Second Circuit in December 1965, and Office of Communication of United Church of Christ v. FCC, decided by the D.C. Circuit in March 1966.

Scenic Hudson concerned Consolidated Edison’s plan to build a pumped-storage hydroelectric power plant along the Hudson River. The at-the-time recently-formed Scenic Hudson Preservation Conference sought to challenge the Federal Power Commission’s licensing decision under the Federal Power Act. Following decades of precedent, the Commission denied them standing. The Commission stated flatly that the conference had no cognizable injury: they had no economic stake, no property rights, no competitive injury. They were merely citizens who liked mountains. In a surprising about-face, the Second Circuit disagreed. Writing for a unanimous panel, the court held that the statutory reference to “recreational purposes” implied that parties who had “exhibited a special interest” in aesthetic and conservational values could qualify as “aggrieved” under the Federal Power Act. It is hard to overstate how substantial of an expansion of standing this was beyond its traditionally understood remit. Prior to Scenic Hudson, your claimed interest had to fall within one of a handful of formalistic, legally protected categories. Afterward, you simply had to enjoy looking at a mountain.

The opinion also contained a procedural innovation. The court found the Commission’s environmental analysis deficient and remanded for further proceedings, with an implicit threat of injunctive relief should the agency proceed before completing a record that the court would find adequate. This combination of broadened standing, judicial review of procedural adequacy, and remand for record development with threat of injunction, contains every building block for what we now think of as NEPA litigation.

Now the story could reasonably end here, but for the second case that supported standing expansion in Data Processing. Three months after Scenic Hudson came down, D.C. Circuit Judge Warren Burger would extend its logic to broadcast regulation. In United Church of Christ, civil rights organizations challenged the FCC’s renewal of a television license for a Jackson, Mississippi station that had egregiously discriminated against black viewers. The Commission denied them standing: they were neither competitors nor parties with economic interests. In his opinion, Burger cited Scenic Hudson for the proposition that non-economic interests could support standing, then went even further. As the “consumers” of broadcasting, he claimed, the listening public had as much right to participate in license proceedings as passengers had to challenge transit fares. Watching television was all you needed to have a right to participate in licensing proceedings.

The similarity in logic and timing between the two cases appears to have been a consequence of a shared intellectual lineage. Among the few sources cited by both opinions on standing is Louis Jaffe’s 1961 article, Standing to Secure Judicial Review: Private Action. Jaffe was a Harvard professor whose 1961 articles on standing explicitly argued for expanding judicial review to “public actions,” that is to say, suits brought by private parties to ensure governmental compliance with law rather than the vindication of any particular private right. In his article, Jaffe suggested that history shows that the Constitution did not require plaintiffs to show invasion of a personal legal interest to qualify as a case or controversy. It is this principle that both Scenic Hudson and UCC would adopt.

After the two decisions vindicated his framework, Jaffe published a follow-up in 1968 in an attempt to fully legitimize the framework. He called such plaintiffs “non-Hohfeldian,” as in standing outside the traditional framework of legal relations; or, more simply, “ideological.” This broad umbrella included nearly all litigants who wished to sue on the grounds that the government had acted unlawfully. Alongside his own addition of the “zone of interests” test, Douglas ratified the framework in Data Processing, in the process citing both circuit decisions for recognizing non-economic injuries. The deed was done; the “injury in fact” test became constitutional doctrine; and the NEPA scourge began in earnest.1

Apologies for the hiatus, the FAI State Permitting effort has been running at full tilt. We have the new edition of the State Permitting Playbook, expanding our coverage to 49 states; an expansion pack focused on what states can do to improve the prospects of nuclear energy in their states; and, thanks to the inimitable Daniel King, a second expansion pack targeting behind-the-meter reforms for states grappling with data centre expansion.

Good at Permitting?

As part of these projects, I have been thinking a great deal about what we mean when we say a state is good at permitting. Good at permitting how? Does the state intrude minimally? Does the state balance environmental and development considerations? Does nothing that would cause an environmental harm ever get built?

If you’re at all interested in the permitting debate—at the state or federal level—you likely have views that adhere to one of the three categories above. But, at the risk of stating the obvious, these are three very different definitions of being good at permitting, with very different yardsticks for measuring success. The first is about providing a minimum level of protection for the worst environmental harms, with a strong focus on development. The second is a middle ground, focused on considering the balance of the upsides and downsides—accepting some harm for some development. The last is skeptical of our ability to perform analysis at all, particularly when there are long-tail risks of ecosystem collapse.

Personally, I fall firmly into the first camp. I believe that humanity’s creations are at least as good and worthy of respect as nature’s, and that nature is more resilient than people give it credit for. With exceptions for areas where the human cost is clear, such as air pollution, or where nature is at its most beautiful, such as the parks system, I believe the state should support new industrial development—not stymie it. However, with that said, a great deal of the furor around permitting is about tradeoffs which do not clearly need to be made. There are other dimensions where permitting can be improved beyond how extensive the legal protections are. But before diving into those waters, first, some concessions.

Allowing Variance

Despite my personal proclivities, some states should obviously have strong environmental protections. For all of the flack the state gets for its protections stifling development, California remains one of the most beautiful and ecologically diverse areas on the planet. From the stark majesty of the Sierra Nevadas to the ancient splendor of the redwood forests to the silent desolation of the Mojave Desert, nearly every inch of the state is speckled with natural wonder.

Compare this with West Texas. I have personally driven through West Texas no fewer than a dozen times. If asked under pain of death, I could not come up with a single notable feature. Perhaps the pumpjacks. Not exactly natural beauty. Quite obviously, these two places should not have the same standard of environmental protection. Whether the baseline is more or less protective, the variance between these places can—and should—remain high.

Fortunately, unlike federal permitting policy, which requires a one-size-fits-all approach, states can tailor their policies to their environment. I find it useful to think about the tradeoffs states have to make in terms of a Pareto curve. A Pareto curve is a simple concept from economics that shows the tradeoff between competing objectives, in this case, between environmental protection and development.

The concept is simple. At the frontier (the curve), for a given level of environmental protection, you expect to receive a given level of development. This tracks intuition. The more restrictions you place on development, the harder it gets to develop anything. Builders are less likely to build in places where there are more hoops to jump through. On one end of the curve is a California, on the other, a West Virginia.

Where a state places itself on this curve is determined by the answers to a series of fundamentally political questions. Is it acceptable to have higher home heating prices in exchange for not building natural gas pipelines? Should new transmission lines be granted rights of way through state forests? Should farmland be allowed to replace wildlife habitat? The answer to these questions depends on the relative value that voters place on these goals.

Getting to the Frontier

These are hard tradeoffs that require careful consideration, but fortunately—as I alluded to above—very few of them must be made right now. In reality, there is maybe one state (Texas) that is anywhere close to the frontier of efficiency. Most states are so far away from effectively managing their regulatory systems that both development and protection can be substantially improved with minimal environmental sacrifice.

For example, very few states usefully track their permits online. Most are decentralized, with each agency hosting its own portal. These portals rarely take into account permitting office capacity, break down the review process into useful steps, or label which party is responsible for which part of the process. I have only come across two that take into account the relevant information and incentives, Virginia and Pennsylvania, and both have reported staggering improvements in their turnaround times. In Virginia’s case, they cut their state permit processing time across the board by 65 percent.

And a permitting dashboard is just the tip of the iceberg.

Care about endangered species? Rather than a consultation process, why not survey industrial corridors to determine which areas should be off-limits and fast-track the rest. Ditto for flood zones and wetlands.

Worried about manufacturing plants skirting emissions limits through use of exemptions? Just place a cap on the entire facility, and use satellites and ambient air monitoring networks to track compliance.

Concerned about illegal water withdrawals or irrigation overuse? Equip canals and wells with flow sensors, and you can track violators in real time. Management technologies have progressed enormously since these state statutes were originally passed, and it is long past time we updated their implementation for the 21st century.

Below, you will find a list of the 7 changes that—in my opinion—have the best cost-benefit tradeoff for improving the process of permitting. The recommendations fit into one of two buckets, the first for capitalizing costs (1-4) and the second for scaling the system up and down (5-7). Both buckets have become feasible because of the massive improvements in information management and monitoring technology over the last 50 years.

1. Data Centralization (Permitting Portals)

This is the big one. The single most important state-level permitting issue is that permits are tracked incredibly poorly. A great deal of the confusion about permitting is downstream of poor data quality. Very rarely do states know which permits are holding up the system, which steps in those permits are the most burdensome, whether it’s the applicant or the office delaying issuance, etc. Unfortunately, not any tracking system will do. To be effective, the tool needs to align the incentives of the agency officials, the applicant, and the state legislators. This is trickier than it sounds. Fortunately, our new guide shows just how to set up such a platform.

2. Group Permits: General Permits, Permit-by-Rule, and Registration Permits

General permits, permit-by-rule, and registration permits all fall under a bucket I call “group permits.” The details differ, but group permits all bundle similar activities and review them in advance before providing a streamlined application process. They have three big benefits. First, for those more concerned about environmental impacts, group permits can economically justify substantially more up front study than could be done on a permit-by-permit basis. Second, by providing clear requirements, group permits provide certainty to businesses. Lastly, they are dramatically faster to process. A list of common group permits can be found in the FAI State Permitting Report at permittingscorecard.com.

3. GIS mapping improvements

GIS (Geographic Information System) mapping improvements are likewise about taking advantage of fixed costs (from recently available, more costly, technologies) to reduce the marginal review time of each individual permit. Instead of running an endangered species survey for each permit, states can run a large, occasionally updated survey highlighting high- and low-risk areas with automatic or expedited approval. The critical habitat is kept safe, and private actors know in advance where they can build. Same goes for stormwater drainage, wetlands, floodplains, historic preservation, and rights-of-way, among others. These are costly systems to set up, but can often be justified by the downstream savings.

4. Cap Permits: Plantwide Applicability Limit Improvements and Flexible Permits

Cap permits such as plantwide applicability limits (PALs) and flexible permits (see Texas’s model) provide hard limits on the outputs that states care about while allowing flexibility in compliance methods. Plantwide applicability limits have this structure for the Clean Air Act, some Clean Water Act discharge permits do as well, as do incidental take permits (endangered species), safe drinking water maximum contaminant levels, fishery permits, green-house gas cap programs, and some hazardous waste permits. These cap permits recognize that technologies and methods of control will change over time, and that businesses should be allowed to meet their obligations in the most cost-effective manner. Unfortunately, many cap permits have shifting, federally-required renewal calculations, which remove some of the certainty for businesses. To the extent that states are able to, they should push to provide certainty about recalculation metrics in advance.

5. Permitting Staff Funding Mechanisms

One of the most common issues with permit approvals is that staff capacity—if it ever comes online—substantially lags increases in permitting demand. In the good scenario, the permitting department is funded by fees that are connected to permit applications. However, even then the hiring and training of extra officials can take months or years. This causes departments to underhire out of fear of future layoffs. And that’s the good outcome. More often, departments are funded by general appropriations, which require waiting for a new appropriations cycle to even start the funding process. Proper funding is a difficult balance to get right, but an ideal system should have a heavy core of baseline appropriations, supplemented by fees-for-service and an auction/performance mechanism with a limited number of fast-track slots—ideally all with retention authority. This preserves the core of expertise while allowing for slower shifts up and down over time, with sharp spikes in capacity if it becomes worthwhile, as evidenced by the auction mechanism.

6. Third-Party and Self-Certification

Another option to alleviate the burden of cyclical permitting demand is to allow for third party or self-certification. Under that system, instead of working directly with the agency, the state or agency sets a neutral standard. Individuals who meet that standard (usually some certification training requirement), can be contracted on the company’s dime to perform the analysis that the state otherwise would. Then, the state simply approves, denies, or requests more information on the submission. A number of states have passed third-party or self-certification legislation, with Tennessee’s being the best in my opinion. The incentives here can be difficult, so certification and conflict-of-interest (or insurance) requirements are critical, as well as a hard appealable deadline on the reviewing agency.

7. Small Tweaks: Parallel Reviews, Single Intake Portals, and Standardized Templates

Parallel reviews are what they sound like. To the extent possible, agencies should coordinate on permits to maximize throughput. If an NPDES permit, a Title V permit, and a wetlands permit are all required, or, more commonly, if multiple separate agencies are required to act on the same permit, these reviews should be conducted simultaneously wherever possible. A single intake portal for environmental permits reduces confusion for applicants by letting them know exactly where to go. Standardized templates likewise reduce confusion for both the applicant and the reviewer by reducing turnaround time. These three may not be the most complicated reforms, but they fix some of the most common pain points for applicants.

Conclusion

Where once it would have been borderline impossible to contemplate a centralized dashboard or parallel departmental review, technology now allows us to solve these coordination issues. Even more than that, it lets us solve them cost-effectively. Now it’s up to us to build them.

Subscribe now

In my previous analysis, I explained how NEPA—the law requiring environmental reviews—evolved from simple disclosure rule to monstrous veto point. Today's question is, what happens to that same process when government agencies have access to AI that can generate unlimited bureaucratic text? It turns out, maybe not as much as one would anticipate. While AI could revolutionize how agencies write environmental reports, transformation will likely happen outside the government.1 Examining the impact of AI on the most time consuming parts of the process—drafting reports, processing public comments, and defending decisions in court—reveals why. Together, they show a private sector set to capture serious benefits, and a public sector set to become a punching bag.2 

A Brief Overview of the NEPA Process

To understand how these models will affect the NEPA permitting process, it is useful to understand the process itself. An exceedingly condensed version of the process (in chronological order) follows:

1. Proposed Action 

2. Environmental Impact Statement (EIS) Determination

3. Project Timetable + Agency Coordination

4. Notice of Intent (NOI)

5. Scoping/Preparing Draft EIS

6. Public Comment

7. Prepare Final EIS (respond to comments)

8. File Final EIS

9. Record of Decision (ROD)

10. Mitigation & Monitoring Implementation

11. Supplemental EIS (if triggered)

12. EPA/CEQ Dispute Resolution (if needed)3

13. Judicial Review

Of all these steps, the three bolded consume the most time. 

The EIS process starts when an agency determines that a federal action—issuing a permit, releasing funds, building a dam—might significantly affect the environment. The agency then spends years drafting a comprehensive report. Once complete, the draft goes public. Citizens submit comments. Then, the agency responds to each substantive comment and revises the report accordingly. Only after can it issue a ROD approving or denying the project. 

But approval is not the end. For six years after the ROD, anyone with standing can sue to block the project. As long as plaintiffs clear the initial pleading hurdle, the project typically freezes while courts review thousands of pages of documentation. After several years acquiring evidence the court will rule the report insufficient and vacate the decision or declare it complete, at which point the aggrieved party may appeal the decision.

Within the process, two tasks consume the vast majority of time and effort: drafting and commenting. After approval, judicial review stands as the highest hurdle. In examining how AI is likely to change each, there are two yardsticks by which to measure: Do environmental protections improve? Does approval come faster? The ideal outcome achieves better environmental protection in less time. The nightmare scenario delivers neither; imagine a dam that takes 45 years to approve and still kills 1000 grizzly bears.

Drafting

To apply AI, it helps to break down drafting into subtasks. The core of each report is four sections with attached appendices. These sections answer four questions: why is this being suggested (the purpose and need statement)?, what the proposed action will be compared to (alternatives analysis)?, what are the resources in question that may be affected (affected environment)?, and what will be the impact on each resource for each alternative (environmental consequences)? Appendices contain data and modeling assumptions that support the assertions made in the other four sections. 

Writing these reports is where the models can be most straightforwardly useful. They are exceedingly long, often clock in at more than 1000 pages (including appendices), and are neither well-written nor rigorously analytical. To give you a taste, here is a randomly selected passage:

The water resources objective was developed to ensure the LTEMP does not affect fulfillment of water delivery obligations to the communities and agriculture that depend on Colorado River water and remains consistent with applicable determinations of annual water release volumes from Glen Canyon Dam made pursuant to the Long-Range Operating Criteria (LROC) for Colorado River Basin Reservoirs, which are currently implemented through the 2007 Interim Guidelines for Lower Basin Shortages and Coordinated Operations for Lake Powell and Lake Mead.

A primary aspect of reservoir operations that potentially affects water resources is related to the monthly distribution of the Lake Powell annual release volume and its resulting impact on reservoir elevations, operating tiers, and annual release volumes. Changes to monthly release volumes have the potential to, in critical time periods, affect reservoir elevations for operating tier determinations, which could in rare circumstances affect annual release volumes.

This style of writing may be sleep-inducing, but LLMs excel at producing it. Further, I suspect when faced with the question, very few people would argue that a human writing 2,000 pages of this schlock for a report no one will read is a good use of taxpayer dollars. The existence of a written artifact justifying government action may be prescribed, but its method of production need not be. As long as a human honestly confirms that the text is a faithful representation of the work undertaken, AI-generated reports are a pure good. Implementation could plausibly save a year of full-time writing work per EIS.4

LLMs could also solve the agency coordination issues that currently add months—and in some cases years—to the NEPA process. Currently, agencies waste enormous time discovering which permits they need, which laws apply, and which agencies must sign off. An AI system could change this overnight. 

Picture it: You’re an official who has been asked to plan a dam on the Colorado River. Instead of scouring the library, you start by querying a database containing thousands of past environmental reviews.5 Then, using the details of the particular project, you submit a query to whittle down the list of potential alternatives. Within minutes, it suggests which project alternatives minimize environmental harm based on what worked before. After settling on the appropriate alternatives, the model indicates which effects can be analyzed using existing data and which will require time-intensive fieldwork. Then it works with you to propose a timeline of the required fieldwork—with fallbacks for unexpected hiccups. This back-and-forth could cut the months of typical prep work to weeks, and substantially reduce the number of unanticipated delays (e.g., missing a critical breeding season) that hold up publication.

This is all achievable with current technology, but is far from the likeliest outcome. Without training and clear guidance on the use of AI, most bureaucrats will shy away from taking advantage of the technology, and those who do will be selected for using it to replace their work, rather than augment it. For project-sponsor-led documents, these fears are less well-founded, but the coordination problem—spending the necessary funds to structure the EPA’s collection of EISs into a useful format—remains.6 Until that hurdle is overcome, both agencies and sponsors are likely to see only small improvements in drafting. 

Measuring with the two yardsticks, the likeliest outcome (i.e., no guidance with bureaucrats doing on the side) is a slight decrease in the quality of environmental analysis (due to the lack of double-checking), with little to no decrease in review time for agencies. Project-sponsored reviews are more likely to speed up significantly, but given the incentives will see little to no improvement in environmental protections. If the hurdles of proper guidance, training, and database creation are overcome, both types of reviews will speed up while strengthening environmental review.

Public Comment

The public comment sections of NEPA review add an additional wrinkle: two-party AI use. 

First, the positives. Today, meaningful participation in environmental reviews requires background knowledge that can leave out locals who may have critical information. For example, a farmer worried about increasing salinity in their irrigation water might know the danger intimately but lack the expertise to write a technical comment that agencies take seriously. AI changes this. That farmer can now describe the problem in plain language, and AI will translate it into the precise, citation-heavy format that gets attention. AI, if allowed to do so, could give agencies the ability to automatically sort through the corpus of submitted comments and retrieve only the most relevant. If done in good faith, AI will reduce the transaction costs of sharing relevant information on both sides and improve the quality of projects constructed. 

Unfortunately, if there is one thing the notice-and-comment process is not known for, it’s good faith. An actual notice-and-comment process typically consists of: a flood of completely useless comments from busybodies, a handful of detailed comments from the relevant lobbying groups (which can be helpful but are incredibly skewed), a dozen or so claims that the entire ecosystem will collapse if the project is approved (see: the Sierra Club or the Center for Biological Diversity) and 3–5 relevant, useful comments from citizens. 

With AI, this ratio is likely to get worse. Now, if an individual wishes to see a project stopped, they can simply input the draft proposal or EIS into an AI and ask it to find potential holes in the analysis. It’s worth remembering that the criticisms need not actually be reasonable and substantive, they simply need to seem so. Prior heuristics indicating substance such as proper formatting, and citations will no longer serve their function. In the future, a flood of thousands of irrelevant—but seemingly substantive—comments generated by AI will swamp the handful of relevant comments that could improve the project. 

Unlike during the drafting process, agencies will be forced to respond to AI notice-and-comment. The burden of response will be too high not to.7 I see two possibilities based on the scale of the flood. If the increase in substantive-looking comments is more than an order of magnitude, agencies will be forced to adopt LLMs to group/respond to comments out of necessity. If the increase is smaller, they may simply plug along, leaving only private project sponsors using LLMs to improve their response speed.

Regardless of which path is taken, the impact of AI on the public comment process is overwhelmingly likely to be negative. Increased accessibility for useful commenters is almost certain to be outweighed by AI slop. The notice-and-comment system seems destined to become a kabuki theatre of fake AI-generated concerns responded to by rote AI-generated responses, with no useful information transferred between parties. Much like a Pacific Island cargo cult, it will retain form but not purpose. Whether this devolution increases or decreases decision timelines will depend on the extent to which agencies are allowed to use these technologies, and the discretion granted to their responses by the courts. All in all, not great.

Judicial Review

Finally, we get to the judicial review process. Fortunately, NEPA judicial review is so poorly structured that it would be difficult to do worse. As with public comment, there will be two countervailing forces. On one side, agencies and sponsors will have the ability to red-team their EISs and ensure that no obvious litigation attack angle exists. A few hours querying the executive summary should reveal any clear missing components (at least to the extent they are included in regulations and not case law). As a result, EISs will become more legally sound. On the other hand, the cost of EIS analysis to draft a complaint will decrease substantially. The ease of spotting oversights will force drafters to red-team with AI, lest the EIS be held up in court by a cheaply drafted complaint.8

For the above considerations, I expect AI to improve judicial review the most, increasing speed and environmental protection. Smart agencies will fix these identified red-team gaps, including gathering better field data on actual environmental impacts, making reports and protections stronger.

The Supreme Court recently handed agencies another advantage. In the Seven County decision, the Court ruled that agencies have broad discretion to decide what belongs in an environmental review. Legal deference in combination with AI-powered red-teaming will leave lawsuits facing much steeper odds. Judges will see fewer valid complaints because agencies will have already addressed them.

This new equilibrium will be highly dependent on adoption by the relevant parties: agencies, sponsors, and plaintiffs. To the extent that any fail to effectively use AI (or are disallowed in the case of agencies), they will be heavily punished. On net, with full utilization by all three, the balance is likely to shift in favor of agencies and sponsors.

Conclusion

On the whole, I anticipate that LLMs will increase the speed of review while maintaining approximately the current level of environmental protection. My uncertainty is highest around the notice-and-comment process, where I suspect the impact will be largest. Use of LLMs in agency-led reviews and to bullet-proof EISs, when combined with the Seven County decision, should substantially decrease the number of lawsuits agencies lose over time. 

Additionally, I suspect there will be a substantial shift from agency to sponsor-led reviews. As the gap in LLM adoption between the public and private sectors grows, the cost and speed differentials will as well. Even with required agency oversight, the ability of private actors to coordinate, draft, and bulletproof environmental reviews will buy a meaningful amount of time. The extent of this shift will be determined by government LLM implementation processes in the coming years. 

Already, sponsor-led EISs are somewhere between 10 and 40 percent faster than those led by agencies.9 Given that sponsor-led projects are selected to be large and complicated (thus able to bear the burden of expensive environmental consultants), I expect the average complexity of sponsor-led reviews to decrease as lower costs open the market to previous marginal projects. This will enhance the perceived gap in speed between sponsor and agency led projects, and drive further privatization. Without a serious focus on improving state AI capacity, private projects will zip through reviews while public works and the communities that rely on them sit stranded in gridlock.

Subscribe now

Briefly, the law requires that agencies consider the potential environmental effects of their actions. It started off simply, with reports sometimes only dozens of pages, but has morphed over time into a monstrosity requiring half-decade-long reviews and multi-thousand-page tomes. Commentators largely agree that the main roadblock is judicial review of agency decisions. But little attention has been paid to how that roadblock got there. By what mechanism did a simple requirement to consider potential environmental impacts turn into one of the most burdensome pieces of legislation in US history?

A Very Brief History of NEPA

Environmental disasters of the 1960s—burning rivers, choking smog, vanishing species—drove Congress to pass NEPA in 1969, requiring federal agencies to produce "detailed statements" assessing environmental impacts before acting. Agencies initially ignored the law: the Trans-Alaska pipeline's first statement ran eight pages for an $8 billion project. When Friends of the Earth sued in 1970, construction halted for five years while lawyers logged 4,500 hours (worth millions today) producing a nine-volume, 3,500-page environmental impact statement. The Supreme Court ruled in 1975 that plaintiffs couldn't recover these massive legal costs, so agencies kept flouting NEPA until the Council on Environmental Quality (CEQ)'s 1976 report documented this widespread non-compliance, prompting Carter's 1977 executive order granting CEQ binding authority over all agencies. Congress then passed the 1980 Equal Access to Justice Act, overturning the Court by allowing fee recovery in suits against federal agencies.1

Examining Party Behavior

To understand why these reviews spiraled so quickly, you have to examine the risk (and therefore cost) allocation. Allocating legal costs is always a tricky task, with a trade-off between allowing the prevailing party to recover its costs and trying to avoid the societal costs of a trial. The British system prefers the former (the “loser-pays” rule), the American system the latter (to each his own costs). The failure modes of each system differ, but they by and large both work well. The EAJA, by contrast, allows plaintiffs to recover their fees if they win but does not force them to pay the government’s legal costs if they lose.2 Other suits with similar provisions include the Endangered Species Act, Americans with Disabilities Act, and the Freedom of Information Act, all notorious for litigation abuse.

By refusing to allow the government to pass along its costs if successful, the EAJA encourages exploitative behavior by plaintiffs. Even a successful defense became an expensive proposition. If plaintiffs are able to get over the pleading barrier, a lawsuit can drag the review out by years through the typical discovery/trial/appeal process, regardless of whether the plaintiff had a reasonable case or not. And without the worry of having to bear the cost, plaintiffs are encouraged to abuse costly legal strategies. Proof of this exploitation can be seen in the incredibly high proportion of NEPA cases that are won by the agency. The agency prevails approximately 80 percent of the time, but in almost none of those cases is it allowed to recover its costs. 

This fee-shifting structure would not be nearly so onerous if the pleading bar were not so low. If plaintiffs were required to plead more specifically, at least the courts would be able to discharge their duty to block frivolous cases on the front end. But because of the 1977 executive order, agencies were required to comply fully with every CEQ regulation. This meant any failure to comply with even one clause in the dozens of pages of text was a procedural defect and theoretically met the pleading standard.3 Agencies took one look at their new incentives, and promptly decided to write 7,000 pages anytime they thought there was a chance the action might be challenged. Unfortunately, even these extensive reviews failed to fully protect their decisions. 

However, even the CEQ regulations and EAJA are not the full story. Though they go some ways to explain the rise in NEPA litigation since the 70s, they do not explain how that rise created such onerous review standards. Previously, I have used the term “ratchet statute” to describe a law that turns in only one direction, becoming continuously more burdensome over time.4 I believe NEPA falls into this category. But to understand how a statute like NEPA can “ratchet,” you have to consider the pressures on the individual actors.

Building a Dam under NEPA

Say you are an enterprising individual who would like to build a dam to provide power to a nearby town. To do this, you will need a Section 404 permit from the Army Corps of Engineers (ACE). Because that permit approval counts as a major federal action under NEPA, and a dam is a substantial undertaking, you will need to complete an EIS. Technically this is the agency’s job, but you’ll likely contract with a consultant to write the EIS and submit it for the agency’s approval. Over the course of three years, you work with the consultants and the public to complete the document. You consider the effect on endangered species and cultural resources, downstream flooding, sediment transportation dynamics, baseline habitat conditions, and a reasonable range of alternatives. With the complete document totalling 600 pages, the agency gives its stamp of approval, and you receive the permit.

Unfortunately for you, a local member of the Sierra Club, Bill, is not a fan of the project, and has threatened to sue. You feel confident in your study, however, and press ahead anyway. True to his word, after the permit issuance, Bill claims that the study failed to consider the effects of the possible success of your dam on encouraging further dam construction in the area, leading to foreseeable environmental degradation. In his pre-approval public comment condemning the dam, Bill had not mentioned this claim, despite noting several dozen other objections. Armed with the legal theory that the failure to address this concern was a procedural defect, Bill files in district court. You think the claim is ridiculous, but the court finds that it meets the pleading standard and issues an injunction to halt the construction while it considers the case. At this point it is the Army Corp’s obligation to defend the permit, but working with the local office for the last three years has shown you how short-staffed they are, so you decide to intervene in the case and help the defense.

Two years later, after pleadings, answers, preliminary motions, and discovery, the court finally dismisses Bill’s claim in response to your motion for summary judgement. The project is no longer encumbered, and you build your dam. Victory is yours.

Unbalanced Incentives and Ratcheting

But what if it weren’t? You win, you get your dam, but what happens if Bill wins (e.g., on appeal)? If the court rules in Bill’s favor, it’s not just your dam that has to undertake review of the cumulative effects of potential future dams, it’s every potential dam under the court’s jurisdiction.5 Every time the Bills of the world win a case, another requirement is added to future EISs. 

Even worse, the new requirements expand fractally. Once the need to conduct cumulative review for further dams is established, a reasonable argument can be made that the endangered species area under review should include potential sites for those future dams as well.6 Though this particular example is far-fetched, even if the Bills of the world had a win rate of 1 percent, requirements would still accumulate over time.

And there is absolutely no counterbalance to this process. If you win, the status quo stays exactly where it is. The next dam you try to build will have to uphold all the accumulated standards that applied when permitting the first dam. If you actually wanted to overturn a particular requirement, you would need to do the following: 

1.) Select a building site 

2.) Apply for a permit 

3.) Go through the entire multi-year EIS consulting process, minus the one particular requirement you want to challenge

4.) Persuade the agency to approve the permit, despite the agency's legal obligations under CFR regulations and established judicial precedent—a decision that would expose the approving official to enormous professional liability

5.) (Likely) get sued

6.) Go through the requisite years of pleading, discovery, and motions 

7.) Convince the judge to overrule the prior precedent despite stare decisis, a ruling they may be bound by; and

8.) Survive an appeal with the same issues

Congratulations! If successful in this incredibly costly, long-shot proposition, you have removed a single requirement from the process. The next EIS will be one half-page shorter.  

Clearly, no one in their right mind would do this. 

It is this phenomenon, where a legal win only preserves the status quo, whereas any loss imposes new requirements on everyone, that defines a ratchet statute. Much hay has been made of the particular requirements of NEPA over time, from the heightened compliance requirements to the automatic injunctions to the lowered proof required for standing, but the exact requirements are less important than the structure.7 If cumulative impact analysis weren’t required, something equally bad would take its place.8

Conclusion

Over decades, what was once a reminder for agencies to consider the environmental impact of their actions has become a noose. The Forest Service has estimated that  40 percent of its total work is planning and assessment. As discussions pick up, reformers must take care to avoid replacing the current system with one with an equally toxic logic. One-way tightening must be replaced by a system that permanently balances the interests of those who wish to build and those who seek to preserve the beauty of nature.

And reform can’t come soon enough. As AI starts gaining traction in the legal community, the game theory of these lawsuits is likely to change once again, potentially leading to a future of less environmental protections and longer reviews. Next week, I will attempt to apply LLMs to all the steps of NEPA review and give my thoughts on how I think the technology is likely to change the status quo. Stay tuned.

Summary

1. NEPA was passed in response to a genuine inability of administrative agencies to properly consider diffuse environmental costs.

2. For its first decade, NEPA’s costs were counterbalanced by the expense of litigation enforcement and the lack of binding regulation to provide a hook for lawsuits.

3. The passage of the Council on Environmental Quality (CEQ) regulations and the Equal Access to Justice Act (EAJA) removed these safeguards and steeply tilted the legal review process towards plaintiffs.

4. If defendants win the case, they merely preserve the status quo; if plaintiffs win, reviews become more onerous.

5. There is no party with an incentive to push back on bad precedents, so agencies comply regardless of whether the requirement is actually informative.

6. Repeat 4-5 for 55 years, and you get the current status quo.

7. Any reform will need to avoid establishing a similar incentive structure if it hopes to be permanent.

But there’s a particular species of failure that's far more troubling. Sometimes laws don't fizzle, they transform into something their authors never intended, or even sought to avoid. The history of the Davis-Bacon Act illustrates this process.

In 1931, Congress faced a problem. Federal contractors were importing cheap labor from other states, undercutting local workers on government projects. The solution seemed simple: require contractors to pay "prevailing wages" based on local standards. To implement the solution, President Hoover signed the Davis-Bacon Act, only two pages long when it was passed.

Congress wrote two pages. The bureaucrats wrote the rest.

Davis-Bacon is what I call a "ratchet statute," a law that turns in only one direction. Like a mechanical ratchet, it can tighten but never loosen. It can add requirements but never subtract them. It can complexify but never simplify. These statues are choking our political system, and the history of Davis-Bacon reveals how.

To implement the bill, the Department of Labor first had to answer the question: what exactly is a prevailing wage? To find out, it had to survey wages for every trade in every locality. Carpenters in Cleveland, electricians in El Paso, plumbers in Peoria–reasonable enough. But administrative tribunals–the internal judicial systems for executive agencies–ruled that these surveys needed more detail, and so the ratchet tightened. "Power equipment operator" was too vague. New classifications like "power equipment operator – heavy and highways," "power equipment operator – paving and incidental grading," and "power equipment operator – sewer, gas and water line" proliferated.1 With each designation came an additional wage determination. 

Then came the compliance requirements. Contractors had to file weekly payroll reports. Workers had to be interviewed to verify their jobs matched their classifications. Disputes went to hearing officers. Appeals created precedents. Each precedent added procedures, and the ratchet tightened further.

Today, that two-page law has metastasized into hundreds of pages of rules. The Department of Labor maintains hundreds of different work classifications across thousands of locales. Contractors hire consultants just to navigate the paperwork. A recent study found that Davis-Bacon compliance adds roughly 7 percent to federal construction costs due to mismeasurement of wages alone. That’s $21 billion annually in extra expenses that buy nothing but paperwork. Defenders of this complexity are scarce. Ask union members privately, and they'll admit the system is absurd. Ask contractors, and they'll show you binders of meaningless forms. Ask government officials, and they'll acknowledge that many classifications make no sense. Only a few individuals in each category benefit from the system, but for those few it provides their entire livelihood. And so the machine grinds on.

Davis-Bacon is only one example of a class of terrible laws. Most ratchet statutes start similarly. Congress passes a law with good intentions. But then courts interpret it, adding requirements based on vague language. Agencies implement those requirements, creating procedures to avoid lawsuits. Interest groups defend their piece of the complexity. Soon, entire industries exist just to manage compliance. Reform becomes impossible because too many people profit from the system.

50 years of nearly flat energy consumption hid the scale of the problem. But as America races to develop artificial intelligence, we're discovering that these laws have broken our ability to build. A procurement system designed for buying typewriters now controls who can sell quantum computers to the government. Personnel rules created for filing clerks determine who can build neural networks for federal agencies. Environmental reviews meant for interstate highways delay data centres for years. The Davis-Bacon Act itself now requires "prevailing wage" determinations for job categories that barely existed five years ago.

Our competitors face no such constraints. While American contractors spend months determining whether an AI technician counts as "Telecommunications Technician" or "Electrician – Inside Wireman," China builds. The gap between what American technology can do and what the American government can actually buy grows wider every day.

Over the coming months, I will examine these ratchet statutes to understand how they work, why they persist, and what we can do about them. I'll highlight four areas: permitting (how we approve projects), procurement (how we buy things), personnel (how we hire people), and rule-making (how we create regulations).

For each statute, I'll invert the analysis. Instead of asking what the law tries to achieve, I'll examine the incentives it actually creates.

For each statute, I will answer five questions:

• What problem were lawmakers trying to solve?

• How did courts and agencies transform the original intent?

• What perverse incentives emerged from those transformations?

• What new problems have these incentives created?

• How can we fix the original problem and the later mess?

Next week, I'll begin with a precise definition of what makes a statute ratchet as seen through the National Environmental Policy Act (NEPA). We'll see why some laws worsen with time while others remain manageable, and what that teaches us about handling the AI future.

As seen above, the number of lawyers in the US has skyrocketed since 1971.2

US politics is dominated by lawyers and the legal system to an extent that is almost unique in the developed world.3 Our public officials, laws, and society are shaped not just by individual lawyers, but by a system of thought that, through its quirks, constrains what government can and cannot be. Nearly everyone has had a personal experience with the consequences of this mode of thought. Sometimes it's a “Caution: Slippery When Wet” sign on a dry floor. Other times it's a plaintiff abusing the Americans with Disabilities Act to bankrupt a small business. 

My goal with Statutory Alpha is to examine how that system is changing, and how AI technology can reform it for the better. 

Because make no mistake, the system is changing. The future of artificial intelligence may be hazy, but its present abilities are crystal clear. Legal reasoning is about to become effectively free.4 Many critical safeguards holding the system together (e.g., high barriers to entry, the law school training system), will be undermined by AI. Some already have been. What the legal system will look like on the other side of this shift is anyone’s guess, but the problems it will have to adapt to are emerging now. AI-drafted slop legislation, the nuclearization of nuisance lawsuits, and an automated FOIA request flood are all just around the corner. Even nastier unknown problems likely await us. 

Yet for all the disruption AI will bring to our legal system, it will also afford an opportunity to fix many of the system’s worst features. Easy access to legal reasoning could offer justice to millions whom high fees exclude. AI assistants can help those representing themselves navigate court procedure, translate legal jargon into plain English, and level the playing field for those who can’t afford to pay $1000/hr. Contract analysis and arbitration could be automated, driving down costs and making those services accessible to small businesses and individuals. Perhaps most importantly, AI could free us from the scourge of vague legislation and its consequences. Rather than overwhelming the system, AI could empower Americans to once again understand the laws that govern them.

Making the most of this opportunity will require grappling with these issues before they become emergencies. As the old saying goes, “hard cases make bad law,” and emergencies make the worst law of all. It is up to us to determine what good law should be before circumstances force our hand. This blog will be dedicated to exploring that question: How can we adapt our legal system to best take advantage of AI? 

Currently: I am taking a look at a group of laws that I suspect is going to interact particularly poorly with AI—I call them ratchet statutes. First up, the National Environmental Policy Act.

Subscribe now