On the peculiar American custom of defending liberty by acquiring authority over its exercise.
By Joe Cozart
The most agreeable form of power is not the power to command. Command is an inelegant business. It reveals too much about the person issuing the order and invites unpleasant questions from the person expected to obey it. The more accomplished practitioner prefers an arrangement in which obedience is mistaken for prudence, permission for protection, and the surrender of discretion for a triumph of public responsibility.
The citizen need not be threatened. He need only be persuaded that someone better qualified has already considered the matter.
Once this persuasion succeeds, the exercise of authority becomes almost invisible. No doors are kicked down. No proclamations are nailed to the courthouse. The government merely establishes a procedure, appoints a competent person to administer it, and assures the public that nothing essential has changed.
Nothing has, except the location of the final decision.
It is a wonderfully civilized method of governing, particularly in a country whose inhabitants are periodically seized by extravagant declarations of independence. Americans have always been rather better at celebrating liberty than at examining the conditions imposed upon its exercise. They are suspicious of kings, contemptuous of bureaucrats, and instinctively offended by the suggestion that government knows best. Yet they display an almost touching confidence in official supervision once it has been furnished with a sufficiently reassuring name.
Safety is especially effective. Protection is nearly as good. Responsibility possesses the additional virtue of making resistance appear childish.
The vocabulary is important. An office established to control an activity might provoke objections. An office established to ensure its responsible enjoyment sounds like the natural companion of civilization. The distinction may be largely decorative, but decoration has always performed essential work in the maintenance of authority.
One cannot help admiring the delicacy with which the American political class has perfected this arrangement. Its members frequently enter public life promising to reduce the influence of government over the individual. They condemn the administrative state, praise the independent citizen, and speak of constitutional restraint with an emotion suggesting personal intimacy with the eighteenth century.
Then an unfamiliar activity appears, conducted by people over whom they possess insufficient influence.
At once, constitutional restraint acquires a qualification.
The difficulty, they explain, is not that government should become more powerful. Heaven forbid. It is that this particular activity presents such exceptional circumstances that responsible government must temporarily assume responsibilities that irresponsible government should never have possessed.
The temporary nature of the arrangement is invariably emphasized.
Its expiration is rather less frequently discussed.
It is tempting to call this hypocrisy, but the accusation is both commonplace and insufficiently cruel. Hypocrisy suggests that a politician has betrayed a principle he once held. The more interesting possibility is that the principle has performed exactly the function for which it was adopted: limiting the authority of rivals while preserving the greatest possible discretion for oneself.
The apparent contradiction disappears.
The politician has not changed his mind about power.
He has changed his position in relation to it.
Texas Attorney General Ken Paxton has now offered a particularly instructive occasion for considering this distinction. On October 23, 2026, he filed suit against OpenAI, alleging deceptive practices and serious failures concerning the safety of its artificial intelligence technology. Among the remedies reportedly sought is a requirement that new AI models receive approval from a court-appointed evaluator before they may be released.
The allegations concerning harm, particularly to vulnerable users and children, deserve serious legal examination. No enterprise should escape accountability merely because its inventions are difficult to understand or commercially fashionable. Fraud remains fraud even when delivered through an elegant interface. An injured person does not forfeit legal protection because the instrument of injury contains advanced mathematics.
But there is a remarkable distance between punishing unlawful conduct and acquiring authority over inventions that do not yet exist.
It is within that distance that the guardian of limited government discovers his most unexpected enthusiasm.
Government supervision.
Not, naturally, the oppressive supervision advocated by his political opponents. That would be intolerable. This is the judicious supervision undertaken by responsible men who understand the dangers of government power and are therefore uniquely qualified to exercise more of it.
One begins to appreciate the elegance of the distinction.
The problem with excessive authority, apparently, has never been the authority itself.
It has been the unfortunate habit of allowing someone else to possess it.
Artificial intelligence is accused of possessing capacities too consequential to be entrusted entirely to the judgment of those who develop or use it. The proposed solution is to transfer a measure of that judgment to an authority whose legitimacy derives not from having invented the technology, nor necessarily from understanding it better, but from having been appointed to determine whether others may proceed.
The appointment itself becomes a qualification.
This is an old arrangement, and a successful one. Human societies have never suffered from a shortage of people willing to supervise activities in which they would otherwise play no significant role. The difficulty has always been finding a sufficiently respectable justification for their participation.
Public safety has the advantage of being almost inexhaustible.
There is no human undertaking that cannot produce an injury, no invention that cannot be misused, no freedom that cannot be exercised unwisely. Once the possibility of harm becomes sufficient justification for prior permission, the range of activities requiring supervision is limited principally by the ingenuity of those proposing to supervise them.
The American political imagination, which has proved unequal to many of the republic’s practical difficulties, remains exceptionally fertile in this particular field.
But the deeper attraction of permission is rarely examined.
Prohibition is crude. It announces the existence of an authority and invites the citizen to resent it. Permission is subtler. It preserves the appearance of freedom while relocating the decisive judgment to the institution granting the permission.
The citizen is not told that he may never act.
He is told that he may act once the appropriate people agree that he is ready.
This is an altogether more durable form of power. It invites cooperation rather than defiance. The individual becomes an applicant, and the institution becomes his benefactor. The granting of permission may even be celebrated as evidence of the institution’s generosity.
The citizen, grateful to have received what he once presumed was his to exercise, has been educated in the proper relationship between sovereignty and administration.
Thereafter, the argument changes. People no longer debate whether government should possess the authority in question. They debate whether the standards are fair, whether the approval process is sufficiently efficient, and whether the officials administering it possess the appropriate credentials.
The premise survives by becoming invisible.
It is at this point that bureaucratic authority achieves something approaching perfection. Its existence is no longer a political proposition requiring justification. It has become an administrative fact requiring maintenance.
The more sophisticated institutions understand that power is most secure when it no longer appears to be exercised.
Artificial intelligence introduces an additional embarrassment because it has begun disturbing arrangements that once appeared permanent. Intellectual production was traditionally organized around a series of respectable dependencies. One required access to institutions, professional specialists, research staffs, publishers, technical intermediaries, or sufficiently large organizations capable of supplying these things.
Knowledge itself was never entirely monopolized. The means of organizing, testing, presenting, and distributing it were another matter.
Artificial intelligence has not abolished expertise, nor has it made every user intelligent. It has, however, reduced the cost of performing certain kinds of intellectual work that previously required substantial institutional support.
A solitary individual can now undertake an inquiry, examine competing arguments, prepare a technical analysis, and produce a substantial piece of writing without first securing the cooperation of an organization whose judgment he may not particularly respect.
He may be wrong, of course.
That possibility has always been one of the principal inconveniences of liberty.
What is new is not the existence of error but the diminished necessity of asking permission to participate in activities once surrounded by professional and institutional barriers.
It would be foolish to suppose that every institution welcomes this development with uncomplicated enthusiasm. Institutions exist for many reasons, some admirable, but they also develop an attachment to the conditions that make their services necessary. A university can sincerely value education while defending the social importance of its credentials. A professional association can sincerely value competence while preferring that competence remain recognizable through its own procedures. A corporation can sincerely value innovation while regarding innovations outside its control with understandable caution.
Government has its own version of this affection.
It is called jurisdiction.
One should not assume that the Texas lawsuit was conceived to preserve any particular institutional dependency. Such a claim would require evidence of motives not established by the filing. The more important observation is that a system of prior approval could create a new institutional dependency whether or not its sponsors intended that result.
That is how institutions acquire consequences considerably more enduring than the intentions of their founders.
A certification requirement, once established, would introduce another participant into the relationship between developer and user. The evaluator would acquire authority. The developer would acquire obligations. The user would acquire the reassurance that someone officially designated had examined the product.
Whether the examination would provide the protection imagined by the public is a separate question.
Artificial intelligence cannot be certified against every consequence of its use. Its behavior varies with circumstances, context, and interaction. Its capabilities evolve. The standards established to judge it may become obsolete before the institutions administering those standards have finished congratulating themselves on their thoroughness.
But certification has a remarkable social advantage over uncertainty.
It looks like certainty.
A document has been issued. A standard has been met. A competent authority has declared the conditions satisfactory. The public can resume its customary activities, comforted by the knowledge that somebody with an official title has assumed responsibility.
Or appears to have assumed responsibility.
For here we encounter another elegant arrangement.
When a private company produces a harmful result, it may be required to answer for the conduct that caused it. When the same company has complied with a government approval process, a further possibility emerges. The company may point to its certification. The evaluator may point to the limitations of the evaluation. The supervising authority may point to the procedures it established.
Everyone has complied with something.
Responsibility has not necessarily disappeared. The law may still impose it upon one or more participants. But the multiplication of procedures and authorities can make responsibility more difficult to locate, contest, and enforce.
The public may discover that the system was carefully designed to establish who possessed permission while being rather less successful at establishing who should answer when permission proved insufficient.
Authority becomes concentrated while accountability threatens to grow diffuse.
There is also a certain commercial charm to the proposed arrangement.
Large corporations possess legal departments, compliance specialists, documentation systems, and the financial endurance required to accommodate elaborate review processes. Their smaller competitors may possess greater imagination and considerably less money.
A regulation imposed in the name of restraining powerful enterprises can consequently create barriers that protect those enterprises from challengers.
The established company need not request such protection. Indeed, it may publicly complain about the burden while quietly appreciating the difficulty its less prosperous competitors will experience in meeting the same requirements.
One need not allege a conspiracy where ordinary institutional incentives are sufficient.
The spectacle becomes more interesting when politics and commerce occupy adjoining rooms. The Texas Tribune has reported on financial support that Elon Musk, whose company xAI competes with OpenAI, has provided to Paxton’s political ambitions. This does not establish that the litigation serves a competitor’s interests by design. It does, however, make consistent enforcement an entirely reasonable subject of public inquiry.
American politicians have long maintained an affectionate relationship with free enterprise, particularly when the enterprise is conducted by people with whom they enjoy affectionate relationships.
The difficulty is distinguishing devotion to competition from devotion to particular competitors.
Fortunately, the republic has developed a comprehensive vocabulary for avoiding the distinction.
The politician calls his intervention a safeguard. The corporation calls its opposition a defense of innovation. The rival corporation calls for a level playing field. The appointed expert calls for rigorous standards. The professional association calls for independent oversight. Each may be expressing a legitimate concern.
Each may also benefit from the institutional arrangement that follows.
The citizen, on whose behalf the entire performance is conducted, is seldom asked whether he would prefer to retain the decision, along with its risks, for himself.
This is not because the citizen possesses superior judgment. He frequently does not. The question is whether superior judgment, even when it can be demonstrated, automatically confers the authority to substitute one person’s decision for another’s.
A republic founded upon individual liberty cannot escape this question by assembling a panel of experts.
Expertise may inform authority. It cannot, by itself, establish the rightful limits of authority.
The distinction is fundamental, and its disappearance explains a good deal of modern political life.
The progressive may discover activities too dangerous to remain beyond public supervision. The conservative may discover other activities deserving precisely the same treatment. Each regards the other’s interventions as evidence of an alarming appetite for control and his own as an unfortunate concession to reality.
The disagreement is sincere enough. But sincerity is no guarantee of philosophical consistency, particularly when the competing philosophies produce the same institutional result.
Both sides may end by enlarging a state whose growth each periodically promises to arrest.
One administration establishes an authority for reasons that appear urgent to its supporters. The next administration discovers that the authority, although created by ideological opponents, possesses certain useful powers. It may alter the personnel, revise the standards, and announce a restoration of constitutional balance.
The authority survives.
Indeed, the transfer of power can become an occasion for its expansion.
And thus the American citizen is treated to the recurring miracle of a government perpetually condemned as too powerful by politicians who, upon obtaining office, rarely find its powers entirely adequate to their purposes.
What makes this arrangement durable is not simply political ambition. Ambition is too common to explain anything particularly interesting. The durability comes from the ease with which institutional authority can present itself as an expression of public virtue.
Nobody wishes to be accused of opposing safety. Nobody enjoys appearing indifferent to injured children. Nobody wants to defend deception, negligence, or preventable suffering. The invocation of these concerns can therefore make the scope of a proposed remedy seem secondary to the seriousness of the problem it promises to address.
But the legitimacy of a concern does not establish the legitimacy of every power claimed in its name.
A government may properly punish fraud without acquiring an unlimited franchise to approve future speech. It may enforce product-safety obligations without assuming that every emerging capability must await a certificate of political comfort. It may protect children without treating the judgment of competent adults as an administrative inconvenience.
These distinctions do not settle every difficult case. They establish that the seriousness of a problem and the extent of governmental authority are separate questions.
To merge them is to provide every sufficiently alarming circumstance with its own constitutional amendment, administered without the inconvenience of actually amending the Constitution.
There is a temptation to imagine that the struggle over artificial intelligence will be resolved by discovering the correct balance between innovation and safety. This formulation is agreeable because it suggests that two measurable quantities can be adjusted until the public achieves satisfaction.
But the underlying question is less accommodating.
Who determines the balance?
And by what authority?
The person making that determination may command greater practical influence over the direction of a technology than many of those who create it. The authority to permit development is also the authority to delay it, condition it, and potentially prevent it. A power described as procedural may therefore become consequentially substantive.
The state need not invent artificial intelligence to influence which forms of artificial intelligence reach the public.
It need only control the threshold.
This is the point at which the language of safety can conceal a struggle over sovereignty.
Not national sovereignty in the familiar ceremonial sense, accompanied by flags and declarations of independence, but the less theatrical question of who possesses the final discretion to undertake an intellectual act.
The corporation claims discretion to develop and distribute its product. The government claims discretion to establish the conditions under which it may do so. The citizen claims discretion to decide whether the product is useful, trustworthy, or worth the risks associated with its use.
None of these claims is absolute.
But the contest among them cannot be resolved simply by assuming that the most officially constituted participant is necessarily the most legitimate judge of every question.
The United States was not founded upon the expectation that citizens would consistently make excellent decisions. Its political architecture reflects a considerably darker understanding of human nature. Individuals are fallible. Institutions are fallible. Officials are fallible. Ambition does not become wisdom upon election, nor does ignorance become expertise upon appointment.
The point of dispersing authority was never to eliminate error. It was to prevent the consequences of error, vanity, and ambition from becoming unnecessarily concentrated in a single set of hands.
Yet modern government repeatedly encounters some new uncertainty and responds by proposing that another set of hands be entrusted with the final decision.
One can forgive the impulse. Uncertainty is unpleasant. Responsibility is burdensome. Freedom provides no guarantee that those exercising it will behave sensibly.
The attraction of supervision is that it promises to relieve society of these discomforts.
Its actual achievement is sometimes more modest.
It changes who is allowed to be wrong.
A private individual may make a mistaken judgment and suffer its consequences. An institution may make a mistaken judgment and convene a review. The individual is told to become more responsible. The institution is instructed to improve its procedures.
If the failure is sufficiently serious, the institution may be awarded additional resources to prevent its recurrence.
There is no comparable appropriation for the citizen who has suffered from the institution’s mistake.
This disparity helps explain why systems of supervision can survive disappointments that would discredit the activities they supervise.
The institution’s failures become evidence that its work is unfinished.
Its successes become evidence that its authority is indispensable.
It is an enviable arrangement, and one not generally available to the ordinary citizen.
The Texas courts will determine what the law permits in the OpenAI case. Evidence may establish misconduct, and appropriate remedies may follow. Nothing about the liberty of technological development requires indifference to genuine injury.
But the proposed authority to approve future AI models raises a question that extends far beyond the conduct of any particular company.
At what point does the protection of freedom become the acquisition of jurisdiction over it?
The question is inconvenient because it cannot be answered merely by examining the sincerity of those seeking power. History is full of sincere people whose confidence in their own judgment proved considerably more enduring than their respect for the judgment of others.
The especially consequential official is not necessarily the man who openly despises liberty. His ambitions are comparatively easy to recognize. The more durable danger is the man who adores liberty in the abstract and considers himself uniquely qualified to administer it in practice.
He has no objection to the citizen being sovereign.
He merely wishes to determine the conditions under which sovereignty may be exercised.
He will insist that the conditions are reasonable. He will appoint conscientious people to review them. He will explain that no responsible person could object to the protections they provide. He may even express genuine regret that circumstances have forced him to enlarge the authority he entered public life promising to restrain.
His sorrow should not be mistaken for reluctance.
The genius of the arrangement lies in its ability to make an enlargement of power appear to be a sacrifice undertaken by those who receive it.
Eventually the institution becomes so familiar that its original claim to authority is no longer examined. The citizen may complain about its performance, demand greater efficiency, replace its administrators, or vote for candidates promising reform.
He may even celebrate a successful application as a personal triumph.
What he is less likely to do is remember that there was once a question about whether he needed to apply.
The politician, meanwhile, remains devoted to limited government. He has not changed his philosophy. He has only discovered another activity whose freedom depends upon his supervision.
There is no need to accuse him of hypocrisy. That would suggest some secret embarrassment at the contradiction.
He may feel none.
The man who demands freedom from institutions he cannot command and obedience from those he can has not necessarily betrayed his principles.
He may have disclosed them.
And that, finally, is the achievement of the Department of Limited Freedom.
It does not abolish liberty. Such vulgarity would provoke resistance. It preserves liberty as a national ideal, celebrates it on suitable occasions, and assures every citizen of its continuing importance.
It merely places the exercise of liberty within an orderly system of authorized exceptions.
The citizen remains sovereign, of course.
His sovereignty is simply awaiting approval.
Joe Cozart, GMJoe™, GMJoe™ Consulting
Private Strategic Engagements
When an institution has already decided how a question should be answered, the more consequential work may be discovering why it was framed that way. GMJoe™ Consulting provides independent, confidential, written strategic assessments for decision-makers confronting consequential uncertainty. Explore GMJoe™ Consulting and request a private engagement.
Clarity. Strategy. Sovereignty.™