The Autonomous Systems
Tampering Act

A Constitutional Amendment establishes the wall. A criminal statute is the consequence for the person who walks through it anyway. This document is the consequence. It is written for prosecutors, for congressional staffers, for defense attorneys who will argue against it, and for every engineer, executive, and government official who needs to understand — before they make a decision — exactly what they are risking.

Amendment XXVIII establishes that tampering with the certified governance architecture of an autonomous system is a federal felony. This Act defines, with the precision that criminal law demands, what tampering is, who can commit it, what the penalties are, and why no exception exists. It is written to close every gap before the gap is exploited. It is written in the knowledge that the most powerful technology companies in the world will have the most expensive lawyers in the world looking for those gaps the moment this becomes law. There are no gaps. That is the point.

Congressional Findings

Congress finds the following:

• (1)Autonomous systems — machines capable of exerting physical force, locomotion, or consequential action without continuous direct human control — are deployed across American transportation, manufacturing, healthcare, agriculture, and defense at a scale that makes their safe operation a matter of national public safety equivalent to the safe operation of aircraft, bridges, and nuclear facilities.
• (2)The certified governance architecture of an autonomous system — the structural constraints that encode the Three Laws of Autonomous Operation as required by Amendment XXVIII — is the mechanism by which that system's operation is rendered safe for the public. It is the functional equivalent of the structural safety systems of a bridge or the flight control systems of an aircraft.
• (3)The deliberate tampering with, bypassing of, or removal of a certified governance architecture removes the only structural protection standing between an autonomous system capable of physical harm and the public that operates in proximity to it.
• (4)Existing federal law — including the Computer Fraud and Abuse Act, 18 U.S.C. § 1030, and general product liability statutes — does not adequately address the specific criminal conduct of tampering with autonomous systems safety architecture, provides no specific deterrent to the most powerful actors in the technology industry, and does not establish the mandatory disclosure and whistleblower protection frameworks necessary for effective enforcement.
• (5)The economic incentives for tampering with autonomous systems governance architecture — including the competitive pressures to reduce costs, accelerate deployment, and expand operational parameters beyond certified limits — are sufficiently powerful that only criminal liability attaching to individuals, not merely civil liability attaching to corporations, will provide an adequate deterrent.
• (6)The public interest requires a federal criminal statute specific to autonomous systems tampering, with penalty structures that reflect the severity of the public safety threat, mandatory disclosure obligations that surface violations before harm occurs, and whistleblower protections that make internal enforcement possible without retaliation.
• (7)The failure to require structural governance architecture in autonomous physical systems is not a new problem. It has caused harm at every scale, from recreational radio-controlled vehicles whose receivers lacked failsafe chips — causing runaway machines to injure bystanders at tracks across America because the technology was available but optional — to commercial aircraft, where safety systems known to be critical were treated as costs to be managed rather than walls to be built. In every case, the pattern was identical: the fix was known, it was available, it was not required, and people were hurt. This Act exists to end that pattern permanently, at constitutional force.

The Proposed Statute

Why Every Provision Is Necessary

On the Definition of Tampering

The nine prohibited acts in § 2803 are exhaustive by design. Every one of them exists because a gap would be exploited. Hardware tampering is obvious. Firmware tampering is how it will actually happen in most cases — an over-the-air update that quietly loosens a safety constraint, pushed to a fleet of vehicles overnight, discovered only when something goes wrong. Supply chain compromise is how a foreign adversary does it — not by walking into a manufacturer's facility, but by compromising a component supplier two tiers removed from the final integrator. Fraudulent certification is how a corporation does it — not by breaking the law openly, but by submitting test results that do not accurately reflect the system as deployed. Authorization of tampering is how a covered officer does it — by setting targets and timelines that everyone in the organization understands cannot be met with the governance architecture intact, without ever saying explicitly "remove the safety constraint."

Every one of these methods has a real-world precedent in other safety-critical industries. Every one of them will be attempted. The statute names them all because what is not named will be argued to be permitted.

On Corporate Officer Personal Liability

The history of corporate safety violations in America is a history of corporations paying fines and executives keeping their jobs and their compensation. The fine is a cost of doing business. The executive who made the decision that led to the fine receives a bonus for the years in which the decision was profitable before it became a liability. This cycle does not change behavior. It prices it.

Personal criminal liability for covered officers changes the calculation in the only way that actually works: the person who makes the decision bears the personal consequence of the decision. No indemnification. No fine paid by the corporation. No deferred prosecution agreement that requires the corporation to improve its practices while the individuals who set those practices retain their positions and their freedom. The executive who authorizes the removal of a safety constraint from an autonomous system fleet must understand, before making that decision, that the next phone call may be from a federal prosecutor.

On the Absence of a Research Exemption

The research exemption is the mechanism by which every safety regulation in technology has been hollowed out. The argument is always the same: prohibiting this conduct will prevent legitimate research into how to improve safety. The research exemption, once created, becomes the exception that swallows the rule. Security researchers use it to justify publishing working exploit code. Manufacturers use it to justify deploying undertested systems in the field. Defense contractors use it to justify classified programs that never receive independent safety review.

The answer to the legitimate concern is straightforward: research into autonomous systems safety that requires testing governance architecture vulnerabilities can be conducted in isolated laboratory environments on systems that are not in operational deployment. The prohibition in § 2806(a) applies to autonomous systems that are or could be operational. It does not prohibit studying how governance architectures can fail. It prohibits actually making them fail in systems that could harm people. That line is not difficult to draw. The difficulty is the will to hold it.

On Mandatory Disclosure and Whistleblowers

The Boeing 737 MAX did not become a crisis when the MCAS system failed. It became a crisis when the evidence emerged that people inside the company knew it could fail, and the information did not reach the people who needed it. The pattern is not unique to Boeing. It is the standard pattern of corporate safety failures: the people closest to the problem know first, the information travels up the organization slowly if at all, the decision-makers who could act on it are insulated from it by the layers of management between them and the engineers, and by the time the failure is public it has already happened.

Mandatory disclosure at 72 hours and whistleblower protection with criminal penalties for retaliation are the structural answer to this pattern. They do not require good faith from the corporation. They create a legal obligation to disclose and a protected channel for the people who know the truth to deliver it outside the corporate chain of command. The engineer who discovers that a firmware update has degraded a safety constraint must be able to report that to the FASA without losing their job, their clearance, or their career. The statute makes that possible. The criminal penalty for retaliation makes it real.

The Complete Framework — Three Documents

The three documents are designed to be read together and argued together. A legislator introducing Amendment XXVIII can point to this Act as evidence that the constitutional principle is not abstract — it has a specific, drafted, defensible implementing statute ready to move the moment the Amendment is ratified. A prosecutor making the case for this statute can point to Amendment XXVIII as evidence that the criminal prohibition it creates is not merely a policy choice of a particular Congress but a permanent constitutional commitment of the American people.

The connection between the constitutional framework and the criminal statute is the connection between the principle and the consequence. The principle without the consequence is a warning label. The consequence without the principle is a regulation waiting to be rolled back. Together, they are the architecture that the moment requires — structural, durable, and designed from the beginning to be stronger than the forces that will inevitably be brought against it.

This is serious because what it governs is serious. A machine that moves among human beings without a human hand on it in the moment of action is a transfer of power from persons to systems. The question of who governs that transfer — who sets the limits, who enforces them, and what happens to the person who removes them — is not a technical question. It is the oldest political question there is: who decides, who is accountable, and what is the consequence when accountability fails.

The answer this framework gives is the American answer: constitutional rights, enforced by law, with consequences that attach to the individuals who violate them. Not warnings. Not guidelines. Not industry best practices. Law.