AI and the Rising Wall of Legal Complexity

By 2019, the Code of Federal Regulations contained nearly 103 million words and about 1.08 million regulatory restrictions, as measured by RegData using terms such as "shall," "must," "may not," "required," and "prohibited." FederalRegister.gov's page-count statistics show that the 2024 Federal Register reached 106,109 actual pages, the highest annual total in that series. That accumulation predates any question about artificial intelligence.

AI tools have begun to change how individuals interact with that volume. In 2016, The Guardian reported that DoNotPay's creator said the service had taken on 250,000 parking-ticket cases in London and New York and won 160,000 of them.

A broader legal-technology market now promotes AI tools for research, drafting, summarization, and document analysis, compressing work that previously required substantially more manual effort.

That shift raises a structural question. AI lowers the cost of reading dense legal text, but it also lowers the cost of drafting, revising, and cross-referencing it. Whether the net effect is a more legible legal system, a more intricate one, or some combination of both is now a live governance question.

The Scale of Federal Regulation

Prescriptive words in the Code of Federal Regulations, terms such as "shall" and "must," grew from about 400,000 in the 1970s to more than 1.1 million in recent counts, according to analysis cited by the Kenan Institute. That count covers only prescriptive terms; as of 2019, the full CFR text ran to nearly 103 million words across all categories.

Whatever view one takes of individual rules, the cumulative text has expanded over decades.

The Dodd-Frank Wall Street Reform and Consumer Protection Act, published by the U.S. Government Publishing Office, illustrates how statutory length compounds downstream. The public-law PDF runs 849 pages. A statute of that size does not merely state a policy objective; it distributes implementation across definitions, delegated authorities, deadlines, reporting duties, and later rulemaking.

That matters because navigation itself becomes a competitive function when operational content is spread across hundreds of pages and then supplemented by later administrative materials. Organizations with dedicated legal and compliance staff are better positioned to absorb that burden than firms or individuals without them.

When operational content is spread across hundreds of pages, organized interests with concentration, income, and ease of organization can act more effectively in regulatory politics than diffuse groups facing higher coordination costs - a structural point George Stigler formalized in his 1971 analysis of economic regulation in the Bell Journal of Economics and Management Science.

The Persistence of the Justice Gap

The volume and complexity of regulation carry direct consequences for individuals. The Legal Services Corporation's 2022 Justice Gap Study found that low-income Americans do not get any or enough legal help for 92 percent of their substantial civil legal problems. The same summary states that about 50 million Americans live in households below 125 percent of the poverty threshold.

LSC's summary also notes that low-income Americans seek legal help for only one out of every four substantial civil legal problems, that cost is a major barrier, and that many people are unsure whether they could find and afford a lawyer if they needed one.

Complexity raises the value of specialized expertise and therefore raises the effective barrier facing anyone without routine access to it.

AI Tools and the Accessibility Question

AI tools have begun to reduce some of that disparity at the margin. The DoNotPay example showed that software could help users contest procedural decisions at large volume. Commercial legal platforms now market AI tools for legal research, brief or memo drafting, contract drafting, summarization, and document analysis.

That does not make them equivalent products, but it does show that AI-assisted legal work has moved beyond novelty.

For individuals and smaller organizations, these tools can lower the threshold for initial comprehension. They can summarize dense statutory or contractual language, surface clauses that may require professional attention, and shorten the time needed to get oriented in a document set.

Even so, the tools do not remove procedural hurdles, guarantee legal outcomes, or replace licensed counsel where counsel is needed.

Whether those gains persist is a separate question. Access to interpretation tools does not automatically change the underlying distribution of money, staffing, institutional familiarity, or political influence that helped produce the justice gap in the first place.

An Analogy from Cryptographic History

Cryptography offers a useful analogy, though not a direct precedent. NIST's guidance on cryptographic transitions is built around a simple reality: algorithm breaks and more powerful computing techniques force transitions to stronger keys and more robust algorithms. In that domain, better tools for analysis do not freeze the system in place; they create pressure for more resilient designs.

Law is not encryption, and regulatory drafting is not adversarial cryptanalysis. But the incentive pattern is recognizable. If AI lowers the cost of parsing dense statutes and regulations, actors who benefit from opacity may also find it cheaper to produce denser, more heavily cross-referenced text.

That incentive does not require central coordination. It can arise wherever complexity carries strategic value.

The asymmetry also matters. In legal systems, the parties best positioned to draft or shape complex text, including large institutions, trade groups, and well-resourced legal departments, are often also the parties best positioned to navigate it afterward. AI extends more parsing capacity to smaller actors, but it also gives incumbents cheaper drafting assistance.

Generative AI and the Cost of Drafting

Generative AI is already marketed as a drafting assistant for legal work. Thomson Reuters describes use cases that include legal research, brief and memo drafting, contract drafting, and correspondence drafting. LexisNexis describes Lexis+ with Protégé as a legal AI workflow solution for drafting, research, and analysis, with features that include summarization, document analysis, and full-document drafting.

That reduction in drafting cost applies to clarifying language and to complicating it. The same systems that can help compress a contract summary can also help generate longer drafts, more clauses, more variants, and more cross-references.

If drafting friction declines, one practical brake on unnecessary complexity declines with it.

The result is not predetermined. AI may still improve access on net, especially where the prior baseline was no meaningful assistance at all. But it does mean that improved legibility for readers and increased productive capacity for drafters will arrive together, not sequentially.

Structural Responses

Several categories of response follow from these dynamics. One is better publication infrastructure. FederalRegister.gov already provides machine-readable statistics about page counts and document categories. Extending comparable discipline to how statutes, regulations, amendments, and cross-references are published would make complexity easier to track over time and easier to compare across domains.

A second is clearer drafting. Plain-language requirements are usually discussed in relation to public-facing explanations, but the same concern applies upstream. Where the legal text itself is unnecessarily opaque, readability becomes more than a communications issue; it becomes part of the allocation of practical power inside the regulatory system.

A third is public-interest legal AI. LSC's figures show that the underlying access problem is already large. Publicly funded legal AI tools, maintained as civic infrastructure rather than sold only as enterprise products, could provide a baseline layer of comprehension for people who currently lack access to counsel.

Such tools would not replace professional judgment, but they could lower the threshold for recognizing when professional help is needed.

The present evidence does not support a simple conclusion that AI will either simplify law or make it less accessible. The stronger claim is narrower. The legal system is already large, already difficult to navigate, and already marked by uneven access to expertise.

AI changes that environment in both directions at once: it widens access to interpretation while also reducing the cost of producing additional complexity.

Sources

• Frank Hawkins Kenan Institute of Private Enterprise. "Up to Code: The Costs of Regulation and Regulatory Uncertainty." Kenan Institute, 2024.
• QuantGov / Mercatus Center. "Quantifying Regulation in US States with State RegData 2.0." Mercatus Center at George Mason University, 2020.
• Federal Register. "Reader Aids :: Page Count By Category Statistics." FederalRegister.gov, accessed 2026.
• Legal Services Corporation. "Executive Summary." The Justice Gap Report, accessed 2026.
• Samuel Gibbs. "Chatbot lawyer overturns 160,000 parking tickets in London and New York." The Guardian, 2016.
• U.S. Government Publishing Office. "Dodd-Frank Wall Street Reform and Consumer Protection Act." Public Law 111–203, 2010.
• George J. Stigler. "The Theory of Economic Regulation." Bell Journal of Economics and Management Science, 1971.
• NIST. "SP 800-131A Rev. 3, Transitioning the Use of Cryptographic Algorithms and Key Lengths." National Institute of Standards and Technology, 2024 draft.
• Thomson Reuters. "Generative AI for legal professionals: Top use cases." Thomson Reuters, 2025.
• LexisNexis. "Lexis+ with Protégé | Legal AI Solution for Drafting & Research." LexisNexis, accessed 2026.