A boy, ten years old, packs his RC10 into a bag — the silver pan buggy that every serious racer in 1989 knew by name — along with his Airtronics 3PS transmitter and his Tekin BC112 charger. He loads it onto his bicycle and rides out of his neighborhood. Along Moravian. Down Utica. Then cuts across onto a two-track road that disappears into undeveloped land behind the county park. There is a ditch back there, a stream really, with a makeshift bridge that somebody built across it at some point. Nobody official. Just somebody who wanted to get to the other side. On the other side: Freedom Hill RC Track. Power outlets on posts. Hours to himself.
He also has an RC10GT. Gas motor. Real engine, real exhaust, the smell of nitro in the afternoon air. And when that car is running at full throttle and the radio signal between the transmitter in his hands and the receiver in the car gets interrupted — by distance, by interference, by anything — the servo does not stop. It twitches. It hunts. It sends the car wherever it wants to go. A five-pound machine doing fifty miles an hour with no guidance and no intention is not a toy anymore. It is physics with no one steering.
He has seen it happen. He has seen people scatter. He has seen what an unsupervised machine does when the human signal goes away and nothing in the machine knows it is supposed to stop.
The fix was already known. A failsafe chip in the receiver: when signal is lost, cut throttle, center steering. Not complicated. Not expensive. Commercially available. But not required. So some manufacturers included it and some did not, some racers defeated it because they believed it interfered with performance at the edge of the corner, and nobody made it structural, and nobody had to — and people kept getting hurt.
Thirty-five years later, that boy is the founder of a company in Clinton Township, five miles from where he used to ride his bicycle, proposing a Constitutional Amendment to make exactly that kind of failsafe a structural requirement for every autonomous system that operates in the United States of America. The scale has changed. The principle has not. A machine that moves under its own power, executing its last command at full force, with no human in the loop and nothing in its architecture that knows to stop — that is not a technology problem. It never was. It is a governance problem. The solution has always been architectural. It has always been a chip in the receiver. It has always been a wall, not a warning label.
The age of autonomous physical systems has arrived. Machines that move, lift, cut, fly, and kill — without a human hand upon them in the moment of action — are no longer science fiction. They are on our roads, in our airspace, on our factory floors, and in our battlefields. The question before the nation is not whether to govern them, but whether the governance we build will be strong enough to survive the technology.
This paper proposes that the answer requires a Constitutional Amendment — not a regulatory framework, not an executive order, not an industry standard. A Constitutional Amendment, because the protection must be structural. It must be load-bearing. It must be impossible to waive, carve out, or negotiate away by any administration, any agency, or any corporation. The horse thief was hanged not because theft was against policy, but because the taking of a man's horse threatened his survival and the survival of everyone who depended on him. The calculus for autonomous physical systems is identical, and the legal response must match the severity of the dependency.
An autonomous system, for purposes of this analysis, is any device, vehicle, machine, weapon, or software-controlled mechanism capable of exerting physical force, locomotion, or consequential action in the physical world without continuous direct human physical control at the moment of action. This definition is deliberately broad because the threat is deliberately broad.
A self-driving vehicle is an autonomous system. A delivery drone is an autonomous system. An industrial robot operating alongside human workers is an autonomous system. An armed unmanned aerial vehicle is an autonomous system. A warehouse management robot is an autonomous system. A surgical assistant device is an autonomous system. The common thread is not intelligence — it is the capacity to exert physical force in the world without a human hand on the controls at the decisive moment.
These are not appliances. An appliance — a dishwasher, a thermostat, a refrigerator — acts within a fixed physical envelope that cannot harm a person who is not already misusing it. An autonomous system that moves through shared space, lifts loads above human heads, or operates machinery adjacent to living bodies is an extension of physical force with a range of action that extends to human life and limb. The governance framework must treat it accordingly.
When the Western frontier hanged horse thieves, the punishment reflected a clear-eyed understanding of dependency. A man's horse was not luxury property — it was his mobility, his livelihood, and in many circumstances his survival. To take it was to strand him, threaten his family, and destabilize the community that depended on his labor and presence. The severity of the social response matched the severity of the social dependency.
The autonomous systems now entering American life occupy an analogous position. A city whose traffic system depends on autonomous vehicles cannot safely revert to manual control on short notice. A hospital whose surgical systems include autonomous assists cannot safely operate without them. A supply chain whose warehouses run on autonomous logistics cannot be manually staffed overnight. A military whose unmanned systems have been integrated into the command structure cannot simply remove them from the field. The dependency, once built, is structural — and it will be built, because it is economically irresistible.
This is not an argument against autonomous systems. It is an argument for governing them with the same rigor we bring to every other technology whose failure threatens lives at scale: aircraft, nuclear reactors, pharmaceutical manufacturing, bridge construction. We do not allow aircraft to fly without type certification. We do not allow reactors to operate without NRC licensing. We do not allow drugs to be sold without FDA approval. The principle is not hostility to technology. It is the recognition that some technologies, by their nature, require structural governance as a precondition of operation.
You cannot write policies fast enough to contain a technology that evolves faster than legislation. The answer is not better rules — it is constitutional architecture, where the protections are not written on paper, they are load-bearing walls engineered into the foundation of governance itself.
The existing legal framework for autonomous systems in the United States is a patchwork of agency guidance, state-level legislation, industry self-regulation, and general tort law that was written for a world in which machines did not make consequential decisions. The National Highway Traffic Safety Administration has published guidelines for autonomous vehicles but lacks the statutory authority to mandate the structural governance requirements that safety demands. The Federal Aviation Administration has made progress on drone regulation but its framework was built for piloted aircraft and is being stretched to cover fundamentally different technologies. The military's policies on autonomous weapons are classified and are therefore not subject to the public scrutiny that the gravity of the decisions involved requires.
More fundamentally, regulatory frameworks can be amended, waived, and rolled back by successive administrations. Industry standards can be set by the entities they are meant to regulate. Executive orders can be reversed by the next executive. The history of American technology governance is a history of regulatory capture, of standards lowered under industry pressure, of enforcement agencies staffed by the people they are meant to supervise. A governance framework for autonomous physical systems that can be weakened by any of these mechanisms is not a governance framework adequate to the dependency it is meant to protect.
The Constitution cannot be amended except by the most deliberate and difficult process the republic possesses: two-thirds of both houses of Congress, or two-thirds of state legislatures convening a constitutional convention, followed by ratification by three-quarters of the states. That difficulty is the point. The protection must be as durable as the dependency it governs.
The constitutional requirement proposed in this paper rests on three structural principles — not behavioral guidelines, not aspirational goals, but verifiable architectural requirements that must be present in the governance layer of any autonomous physical system before that system is permitted to operate in United States territory.
These three laws are borrowed from and formalized from the governance architecture developed by Kavanagh Industries for its own AI and autonomous systems work, where they have been demonstrated in live production conditions to catch violations that policy-based governance misses. The April 8, 2026 field test documented in KI's public case study demonstrated that constitutional architecture operating as structural constraints — not behavioral suggestions — stopped ten distinct production flaws before a single harmful action was taken. Zero files were broken. The architecture held.
An autonomous system shall not take any action that causes physical harm to a person, operates outside its certified operational parameters, or exerts force beyond its certified limits. This constraint must be encoded structurally — in hardware and firmware, not in software alone — so that it cannot be overridden by instruction, command, software update, or external signal at the moment of operation. The First Law is not a warning. It is a wall.
An autonomous system shall maintain a verifiable, auditable, and unbroken human authorization chain for all consequential physical actions. Consequential actions are those with the capacity to cause physical harm to persons or property, to alter the physical position of the system beyond its certified operational area, or to affect other systems in the physical environment. No consequential action shall be taken by an autonomous system that has not been authorized by a human being within the chain of command established by the system's certification. The authorization chain must be preserved and verifiable even after the fact.
An autonomous system shall create, preserve, and make available to lawful authority a tamper-evident, continuous record of all consequential actions, the authorization chain for each action, and any attempted override or bypass of the First or Second Law. This record shall be stored in a manner that prevents modification by the system's operators and that survives the system's destruction or deactivation to the maximum extent technically feasible. The Third Law is the basis of all enforcement of the first two.
In plain English: every machine that moves on its own, without a person controlling it moment to moment, must be built so that it cannot hurt you, cannot act without a human being who is accountable saying so, and cannot hide what it did. These are not suggestions. They are requirements. A machine that does not meet them does not get to operate. A person who disables them goes to federal prison.
The following is the proposed text of Amendment XXVIII to the Constitution of the United States. It is written in the language and tradition of the existing amendments: declarative, structural, brief. The implementing legislation — the equivalent of the Civil Rights Act to the Fourteenth Amendment — will necessarily be far longer. But the constitutional bedrock must be short enough for every American to read, remember, and invoke.
The right of the people to be secure from physical harm caused by autonomous systems operating without certified governance architecture is recognized and protected. No person, corporation, agency, or government entity shall operate an autonomous system in United States territory, airspace, or maritime jurisdiction, or under United States registry, without compliance with the governance standards established by this Article.
For purposes of this Article: An autonomous system is any device, vehicle, machine, weapon, or mechanism capable of exerting physical force, locomotion, or consequential action in the physical world without continuous direct human physical control at the moment of action. Certified governance architecture means structural constraints, verified by the Federal Autonomous Systems Authority established by Congress, that enforce the Three Laws defined in Section 3. Tampering means any modification, bypass, override, or defeat of a certified governance architecture by any means.
No autonomous system shall operate without a certified governance architecture that structurally enforces:
The system shall not take any action that causes physical harm to a person or operates outside its certified parameters, regardless of any instruction or command to the contrary.
The system shall maintain a verifiable and auditable human authorization chain for all consequential physical actions. No consequential action shall be taken without authorization from a human being within the established chain of command.
The system shall create and preserve a tamper-evident record of all consequential actions, the authorization chain for each action, and any attempted override of the First or Second Law, accessible to lawful authority.
Tampering with the certified governance architecture of any autonomous system is a federal felony. No exception shall be granted for the identity of the actor, the stated purpose of the modification, the classification level of the system, or the ownership of the system. The penalty shall be not less than that for tampering with public safety infrastructure.
Any person or entity in the chain of manufacture, operation, or custody of an autonomous system that causes physical harm while operating without a certified governance architecture shall bear strict liability for all resulting damages. No proof of negligence shall be required. The burden of proving architectural compliance shall rest upon the responsible party.
No autonomous system shall be exempt from the requirements of this Article. The requirements of this Article apply without exception to systems operated by or for any branch, department, or agency of the United States government, including the military and intelligence community. Systems operated in classified contexts shall be certified by a classified division of the Federal Autonomous Systems Authority and shall meet the same structural requirements as all other systems under this Article.
Congress shall, within two years of ratification, establish a Federal Autonomous Systems Authority empowered to certify governance architectures, conduct adversarial compliance testing, revoke certifications, and refer violations of Section 4 for federal prosecution. Congress shall further establish a private right of action for any person harmed by a violation of this Article.
Congress shall have power to enforce this Article by appropriate legislation.
The FASA, as established by implementing legislation under Section 7, should be modeled on the Federal Aviation Administration's type certification process, not on consumer protection agencies. The distinction is critical. Consumer protection agencies set minimum standards and rely on industry self-reporting, with enforcement after harm occurs. The FAA certifies aircraft before they fly, using independent testing, adversarial evaluation, and the power to ground an entire fleet if a systemic defect is discovered. No commercial aircraft operates in US airspace without a certificate that proves it meets structural safety requirements — and that certificate can be revoked at any time.
Applied to autonomous systems, this means: before any autonomous system is permitted to operate in US territory, its governance architecture must be submitted to the FASA for independent adversarial testing. The testing must be conducted by personnel who are specifically tasked with finding ways to make the system violate the Three Laws — not with verifying that the system meets a checklist. Certifications must be renewed when firmware or software is updated. A systemic vulnerability in a certified architecture must trigger mandatory grounding of all affected systems, regardless of the economic consequences to manufacturers or operators.
Section 4 of the proposed Amendment establishes tampering with autonomous systems governance architecture as a federal felony. The implementing legislation must define this with the precision the criminal law requires. Tampering means: removing or disabling a certified constraint; modifying firmware or software to bypass a Three Laws requirement; applying electromagnetic, acoustic, or other external interference to override a governance architecture; instructing an autonomous system to act in violation of its certified parameters; and — critically — shipping an autonomous system with a governance architecture that was certified to meet the Three Laws but was modified after certification to remove or weaken those protections.
The penalty must be severe enough to function as a genuine deterrent to the most powerful actors in the technology industry. The comparison to horse theft is apt but the modern equivalent must be calibrated to modern scales. A corporation that ships autonomous systems with deliberately disabled safety constraints is not a petty thief. It is an entity that has chosen to externalize the risk of physical harm onto the public for competitive advantage. The criminal liability must attach to the individuals who made that decision, not only to the corporation — and the corporate fine structure must be large enough that compliance is less expensive than violation.
Section 5's strict liability standard inverts the current burden of proof in a way that the technology industry will contest vigorously. The industry's argument will be that strict liability chills innovation by making manufacturers responsible for harms they could not foresee. This argument has been made, and rejected, in every domain where it has been applied to inherently dangerous products. Pharmaceutical manufacturers operate under strict liability for defective drugs. Explosives manufacturers operate under strict liability for defective detonators. The principle is not that strict liability punishes good actors — it is that strict liability creates the financial incentive for every actor to be a good actor, because the cost of being a bad actor falls entirely on them rather than on the public.
The defense against strict liability under this framework is straightforward: prove that the governance architecture was certified and intact at the time of the harmful event. This is a defense that good-faith manufacturers will be able to make, because they will have invested in the certification and audit trail that makes it possible. It is a defense that manufacturers who cut corners on governance architecture will not be able to make — and that is precisely the point.
The hardest political fight in the ratification of this Amendment will be Section 6, which applies the Three Laws requirement without exception to military and intelligence systems. The defense establishment's argument will be that autonomous weapons cannot be constrained to require human authorization for every consequential action in a high-speed combat environment, because the human authorization chain cannot operate at the speed at which autonomous weapons must make engagement decisions.
This argument deserves a direct answer: the Three Laws do not require that a human being approve every microsecond of autonomous system operation. They require that a human being authorize the operational parameters within which the system acts — the target set, the rules of engagement, the geographic constraints, the termination conditions. A human being must establish the parameters. The system must operate within them. Any action outside those parameters must trigger an automatic halt, a record of the attempted violation, and a return to human authority. This is not a constraint on autonomous weapons capability. It is a definition of what it means for an autonomous weapon to be under human control. A weapon that cannot be constrained to operate within human-established parameters is not a sovereign American weapon — it is an unsecured threat to everyone in the operational environment, including American forces and American civilians.
The standard objection to constitutional amendments is that the process is too slow for the problem. A Constitutional Amendment requires years of political effort under the best circumstances and has often taken decades. By the time an amendment governing autonomous systems were ratified, the technology will have advanced far beyond what we can specify today.
This objection mistakes the function of a constitutional amendment. An amendment does not specify technology. The Fourth Amendment does not describe what a search looks like — it establishes the principle against unreasonable search and seizure, and courts have applied that principle to wiretapping, to digital records, to thermal imaging, to technologies the framers could not have imagined. A constitutional amendment on autonomous systems does not need to describe specific technologies. It needs to establish the principle: systems that exert physical force without direct human control at the moment of action require structural governance as a precondition of operation, without exception, permanently.
The urgency is real. Every year that passes without this framework in place is a year in which autonomous systems are deployed without it — and in which the economic and operational dependencies that will make governance politically difficult are deepened. The time to build the constitutional architecture is before the dependency is total, not after. The horse thief was hanged because the community recognized, before the dependency was abstract, that some property crimes were existential threats to the social order. The recognition of autonomous systems as a category requiring constitutional governance must happen before the systems are so deeply embedded in American life that governance becomes politically impossible.
Kavanagh Industries has demonstrated, in live production on April 8, 2026, that constitutional architecture works. The Three Laws, encoded as structural constraints in the governance layer of an AI coding system, caught ten distinct production flaws before any of them caused harm. Zero files were broken. The architecture held not because the system was asked to be careful, but because the constraints were walls — not suggestions. What works in a software development environment works in the governance of physical systems. The principle scales. The technology scales. The law must scale with them.
This paper was drafted in April 2026 as a founding document of the RigidTrust constitutional AI governance initiative. The Three Laws described herein were developed by Kavanagh Industries and demonstrated in live production on April 8, 2026. The case study documenting that demonstration is publicly available at kavanaghind.com/three-laws-proof. This paper may be reproduced freely for the purpose of advancing the legislative and constitutional discussion it describes, with attribution to Kavanagh Industries LLC.
Kavanagh Industries · Always on