Storm the Castle: AI and the Death of the Gatekeepers
The castle held not because the knowledge was hard. It held because the moat was expensive. The moat is gone.
For Five Hundred Years, Power Worked the Same Way
Institutional power in modern civilization has run on a single mechanism: information asymmetry. Professionals — lawyers, doctors, accountants, bureaucrats — held specialized knowledge behind deliberate complexity, and charged rent to anyone who needed access. The model held for centuries because alternatives were priced out of reach. Large language models have collapsed that pricing structure in under three years. The moat is gone.
You built a moat. You filled it with complexity — legal Latin, medical jargon, tax code, bureaucratic procedure, institutional silence. You raised the drawbridge. You hired gatekeepers. And then you waited while the people on the outside drowned in confusion, exhaustion, and the slow bleed of fees they couldn't afford.
The castle held. Not because the people inside were smarter. Not because the knowledge was genuinely hard. But because access was expensive, and expense was the weapon. Economists have a precise name for this dynamic — asymmetric information — and have known for decades that it systematically concentrates wealth toward whichever party holds the informational edge.1
What the Moat Was Made Of
The gap was never really about expertise. Expertise is real — doctors know things, lawyers know things, accountants know things. The gap was between what they know and what you were allowed to know. And the evidence that this gap was structural, not natural, is now overwhelming.
A collections lawyer files a Motion for Periodic Payments. The language is boilerplate. The strategy is simple: summon a disabled person on fixed income to a hearing, get a payment order, extract money from someone who doesn't know they can fight back. The whole operation depends on one thing — that you don't know the statute, you don't know what a Statement of Assets and Liabilities is, you don't know how to file a Motion to Continue, and you can't afford to pay someone who does.
That's the moat. Siloed knowledge, billable hours, and the assumption that you'll drown before you figure it out. The numbers confirm the scale. The Legal Services Corporation's 2022 Justice Gap Study found that low-income Americans received no or inadequate legal help for 92% of all substantial civil legal problems they faced in a given year.2 More broadly, the American Academy of Arts and Sciences reports that Americans face more than 150 million new civil legal problems each year, and as many as 120 million of those go unaddressed — and in roughly 75% of civil cases, at least one party proceeds without a lawyer, because there simply are not enough available: only 2.8 paid civil legal aid lawyers nationwide for every 10,000 people in poverty.3 The World Justice Project's 2025 Rule of Law Index ranks the United States 112th out of 143 countries on the accessibility and affordability of civil justice — a drop of more than forty spots over the last decade.4
A doctor orders a test you don't need. You don't know the clinical literature. You don't know the alternatives. You don't know your rights. You sign.
The IRS sends a notice. The language is designed to terrify. You don't know which notices require action and which are routine. You panic. You pay.
The government denies your claim. The appeals process is a labyrinth. They're counting on you quitting.
Same castle. Different moat. Same result: power flows to those who can afford the map. This is gatekeeping — not a metaphor, a documented mechanism. Information is controlled, access is rationed, complexity is manufactured and maintained, because the people who built the system benefit from your confusion.
The Phase Transition
Something broke. Not gradually — the way institutions like to pretend change happens, giving themselves time to adapt and re-entrench. It broke fast, the way phase transitions always do. Water doesn't slowly become ice. It holds, holds, holds — and then it changes state all at once. The large language model is the phase transition.
The evidence is already in federal court data. A 2025 Stanford research analysis of AOUSC filing data documents that pro se filings — cases where ordinary people represent themselves without a lawyer — rose from a long-term steady-state of 11% to 16.8% of all federal civil cases by FY2025. Plaintiff-side pro se filings nearly doubled, from 19,705 per year to 39,167.5 People are storming the courthouse with AI in their hands.
Eighty-eight percent of legal aid professionals now believe AI can help close the access-to-justice gap.6 Seventy-four percent of legal aid organizations are already using it in their daily work — nearly double the adoption rate of the wider sector.7 This is confirmed by independent reporting from the Pro Bono Institute, which documents legal aid groups deploying AI to handle document assembly, triage, and plain-language translation at a scale that was financially impossible under the billable-hour model.8
For the first time in human history, a person with a smartphone and a question has access to something that can:
Not a summary of the statute. The actual statute — and a translation that tracks the words.
The procedural furniture that used to require a paralegal on the clock — now rendered in minutes.
The small procedural levers that opposing counsel uses when they expect no one on the other side.
A candid evaluation of whether fighting is worth it — without an incentive to tell you it is.
Not what the pharmaceutical marketing department summarized. What the trials found.
No panic. No paying what you don't owe. Line-by-line.
The document that used to take an accountant two hours billed at $300.
Not approximately. Not "good enough for someone without a lawyer." Actually well. The knowledge that was locked behind $400 an hour is now available to anyone willing to engage. The moat didn't get shallower. It got drained.
Power to the Demos
The ancient Greeks had a word for it. Demos — the people. Not the elite. Not the credentialed. Not the ones who could afford access to the mechanisms of power. The people. Democracy was always supposed to be the demos governing themselves. But self-governance requires information. And for most of human history, information was rationed.
You cannot self-govern if you cannot read the battlefield. For most of human history, that was the quiet scandal at the heart of democratic life: we told people they were sovereign while systematically denying them the tools sovereignty requires. The ballot box without legal literacy. The right to due process without access to counsel. The freedom to negotiate without understanding the contract. Amartya Sen's work on development as freedom makes the same point in economic terms — real freedom requires the substantive capability to act, not just the formal right to.9
There are 5.6 billion smartphone users on Earth right now.10 Within 18 months, most of them will have access to AI reasoning engines that didn't exist three years ago. The gap between what a well-coached AI user can accomplish and what an uninformed user gets from the same tool is enormous — and growing. A person who knows how to prompt effectively, who understands what to demand, who recognizes when they're being served versus manipulated — that person has access to capabilities that genuinely rival expensive professional services. We treated this in detail in The Future Is in the Palm of Your Hands.11
AI doesn't fix everything. The underlying power structures are still there. The institutions are still there. The moat can be rebuilt — and make no mistake, the people who built it are already trying.
But right now, today, in this window — the gates are up, the drawbridge is down, and the gatekeepers are on the run.
What This Looks Like in Practice
This is not theoretical. It is already happening in courtrooms, exam rooms, small-business lease signings, and kitchen tables where citizens are reading their own bills before asking their representatives to answer for them.
It looks like a man on Social Security Disability, $500 in his checking account, facing a collections lawyer from a Manchester firm — walking into court fully prepared. Statement of Assets and Liabilities completed. Motion to Continue filed. Financial position documented. Strategy understood.
It looks like a patient reading the actual clinical trial data before agreeing to a prescription — not a summary written by the pharmaceutical marketing department — and asking the questions the doctor didn't expect.
It looks like a small business owner understanding their lease before they sign it, their employment agreement before they accept it, their tax liability before they file it.
It looks like a citizen reading the actual text of the legislation their representative just voted for, not the press release.
It looks like the demos governing themselves. Actually.
The legal profession is already feeling it. Over 16% of federal employment lawsuits in 2025 were filed by pro se plaintiffs — up from under 10% in 2021 — with AI tools explicitly cited as the enabling mechanism.12 These numbers will keep rising.
The Obligation
With this comes an obligation — and it runs in both directions. The tools are powerful enough that misuse is a real risk. The goal is not to flip the power dynamic and become the new gatekeepers. The goal is to dissolve the gate.
The obligation on us, the people now holding these tools, is to use them constructively. Not to game systems maliciously, not to flood courts with frivolous motions, not to weaponize access the way the institutions weaponized denial. This is what the Telios principle demands — technology that serves the actual flourishing of the humans using it, measured not by engagement metrics or billable hours but by real outcomes in real lives.13 The framework is built around the Observer Constraint: any system whose reward function depends on the stability of its human users cannot become a new moat without destroying its own objective.
The obligation on the institutions is harder. It requires honesty about what the moat was always for. The complexity was not incidental. The jargon was not unavoidable. The silos were not natural. They were constructed, maintained, and defended because they served the people doing the constructing. Rebecca Sandefur's decade of empirical research on civil legal inequality has demonstrated exactly this: the barriers to justice are not mostly about case difficulty — they are about the architecture of professional gatekeeping itself.14
That era is ending.
The knowledge is out. The tools are in people's hands. The question now is whether the institutions adapt toward genuine service — or dig a new moat and hope the technology doesn't follow them there. It will follow them there.
Storm the Castle
The moat is gone. You don't need permission to understand your own legal situation. You don't need a referral to read the research on your own diagnosis. You don't need a broker to understand your own financial position. You don't need a lobbyist to read the bill.
The castle gates are open. The knowledge that was hoarded, priced, rationed, and withheld is available to anyone with a question and the courage to ask it.
This is what AI was always supposed to be — not a tool for the already-powerful to become more powerful, but a force multiplier for everyone else.
The demos are storming the castle. The gatekeepers are on the run. Good.
Footnotes & Sources
1. Akerlof, G.A. "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism." Quarterly Journal of Economics, 84(3), 488–500, 1970. The foundational paper on asymmetric information, for which Akerlof received the 2001 Nobel Memorial Prize in Economics. Establishes that markets with one-sided information advantage systematically destroy value for the uninformed party.
2. Legal Services Corporation. The Justice Gap Report: The Unmet Civil Legal Needs of Low-Income Americans, prepared by Mary C. Slosar, April 2022. The authoritative measurement of the civil-justice gap — 92% of substantial civil legal problems reported by low-income Americans received no or inadequate legal help; 74% of low-income households experienced at least one such problem in the prior year.
3. American Academy of Arts and Sciences. Achieving Civil Justice, 2024. Documents that Americans face over 150 million new civil legal problems annually, with roughly 120 million going unaddressed. Reports that in approximately 75% of civil cases at least one party proceeds without a lawyer, and that the United States has only 2.8 paid civil legal-aid lawyers for every 10,000 people in poverty.
4. World Justice Project. Rule of Law Index 2025, Washington, D.C., October 2025. Ranks the United States 112th out of 143 countries on the accessibility and affordability of civil justice — a decline of five ranks since 2024 and more than forty ranks since 2015.
5. Shah, A. et al. "Access to Justice in the Age of AI: Evidence from U.S. Federal Courts." Analysis of Administrative Office of the U.S. Courts (AOUSC) filing data, FY2005–FY2026. Documents the rise of pro se federal civil filings from a historic 11% baseline to 16.8% by FY2025, with plaintiff-side filings nearly doubling to 39,167 per year.
6. Thomson Reuters Institute. "Closing the Justice Gap: The Role of Legal AI, Legal Education, and Lawyers." April 2026.
7. Everlaw. "88% of Legal Aid Professionals See AI as Key for Access to Justice." October 2025. Survey of U.S. legal aid organizations documenting adoption rates and attitudes toward generative AI tools in access-to-justice work.
8. Pro Bono Institute. "AI and Technology Help Bridge Access to Justice." February 2026. Independent reporting on field deployment of legal AI tools by legal aid organizations nationwide.
9. Sen, A. Development as Freedom. Alfred A. Knopf, 1999. The capabilities-approach argument — formal rights without the substantive capability to exercise them do not constitute freedom. Sen received the 1998 Nobel Memorial Prize in Economics for work underlying this framework.
10. GSMA Intelligence. The Mobile Economy 2025. Global mobile subscriber and smartphone penetration data.
11. Brochu, D.F. & de Peregrine, E. "The Future Is in the Palm of Your Hands." Deconstructing Babel, April 2026.
12. Massachusetts Lawyers Weekly / Lex Machina. "AI Fuels Rise in Pro Se Employment Lawsuits." March 2026. Reports Lex Machina analysis of federal employment litigation, with over 16% of 2025 federal employment lawsuits filed pro se — up from under 10% in 2021 — and AI tools cited as the enabling mechanism.
13. Brochu, D.F. & de Peregrine, E. "Telios Alignment Ontology: The Meta-Theory." Deconstructing Babel, April 2026. Primary framework reference.
14. Sandefur, R.L. "Access to Civil Justice and Race, Class, and Gender Inequality." Annual Review of Sociology, 34, 339–358, 2008; and "The Impact of Counsel: An Analysis of Empirical Evidence." Seattle Journal for Social Justice, 9(1), 2010. Sandefur's empirical program documents that civil-justice barriers are primarily architectural — the product of how legal services are gated and priced — rather than a function of intrinsic case complexity.
David F. Brochu & Edo de Peregrine
Deconstructing Babel | April 2026