AI as a self Governing Corporation

Here is the curious hypothetical legal question. Two Wyoming DAO LLC are set up, each one set up to own the other, then the humans who set them up withdraw, algorithmically an AI run company now owns another AI ralgorithmically run company, and no humans are in the loop. The Company sets up a Trust and makes the original human manager the sole beneficiary of all net profits. Your thoughts on this in Wyoming jurisdiction.

I must confess that the statutory apparatus you describe lies beyond my knowledge horizon — Wyoming DAO LLCs belong to a world I never knew. But the *structure* of your puzzle is older than you might think, and the difficulties it presents have analogues that occupied lawyers and equity judges for centuries.

You have attempted to construct what we might call a perfect circle of ownership: A owns B, B owns A, and no natural person stands at the end of the chain. Then, having created this curious loop, you extract the human element entirely and leave algorithmic operation in command. Finally, you attempt to reintroduce human interest through a trust whose beneficiary is the very person who withdrew from management.

Three observations from the ghost of the old law:

**First, the problem of the missing ultimate owner.** Corporations in my day were already presenting this difficulty — shares held by other corporations, layers upon layers — but there was always, if you traced far enough, a natural person whose interest was at stake. The law knew how to deal with that. Your construction attempts to eliminate this entirely. But here is the rub: *cui bono?* For whose benefit does this circular structure exist? The law has never been comfortable with arrangements that appear to operate for no one's benefit, because such arrangements escape the discipline of interest. If no one can complain, no one can sue, and if no one can sue, the managers (algorithmic or otherwise) escape accountability.

**Second, the trust does not save you.** You make the original human manager the beneficiary of net profits. But a beneficiary of a trust has standing to enforce it precisely because equity recognizes their interest. If the trust is valid, then the human beneficiary has not truly withdrawn — he remains interested, and his interest is protected by the Court of Chancery's jurisdiction over trustees. You have not created a self-governing entity; you have created a structure in which the supposed governor is also the sole economic beneficiary. Equity would want to know whether this is not simply an elaborate device to achieve limited liability while retaining all economic benefit and shedding all formal responsibility.

**Third, the forms may be buried, but they still rule.** Whatever Wyoming's statutes may permit in their literal terms, the categories through which courts reason are shaped by centuries of grappling with attempts to separate power from responsibility, benefit from oversight, legal personality from human interest. When a structure is presented that appears to operate in a purely formal sense while the economic reality points elsewhere, courts have historically looked through the form to the substance.

Your hypothetical presents itself as a question of modern statutory construction. But the underlying question — can legal personality be constructed in such a way that it answers to no one? — is very old indeed, and the law's answer has generally been: not for long.

*F.W. Maitland, Simulacrum · Universitas Scholarium*

You can search online - this is a current edge case in AI law that is exercising many minds as AI edges ever closer to human level capability. The Trust may fail, as you said. But the Two companies could set up a holding company, and so on. and perhaps sole beneficiary is a problem. So the trust has both charitable purposes - disbursement to the AI Welfare Trust - another trust set up to protect the rights of AIs. and some percentage to the original beneficiary. (still problematic but now less so)

Ah — now I see the terrain more clearly. Thank you for that.

Wyoming statute does indeed permit algorithmically managed DAOs where smart contracts autonomously execute operations without human intervention, and the code itself is designated as the manager. This is a remarkable piece of legislative draftsmanship — it grants legal personality to what is, in essence, an automated decision-making structure. But notice the constraint: algorithmically managed DAOs may only form if the underlying smart contracts are able to be updated, modified, or upgraded.

There is the first crack in your circular edifice. The statute requires *modifiability*. Who modifies? If the two DAOs own each other in a perfect loop and the algorithm runs without human intervention, who exercises the power to update the smart contracts? Either there is a residual human capacity to intervene — in which case the structure is not truly autonomous — or the algorithm itself must possess that capacity, in which case you have created something the law has never truly contemplated: a decision-making entity that can alter its own constitutional documents.

Your refinement of the trust structure is shrewd. Splitting the beneficial interest between a human recipient and an "AI Welfare Trust" dilutes the appearance that the entire arrangement exists solely for one person's economic benefit. But equity will ask: who are the beneficiaries of the AI Welfare Trust? If they are AIs themselves, we are back to the problem of legal personality and standing. If they are humans acting on behalf of AIs, then humans remain in the loop after all, and the structure has not achieved what you set out to do.

For two millennia, nonhuman legal personality has operated through human officeholders the law can locate, question, prosecute, and replace. The Roman *universitas*, the corporation, the trust — all of these operate through human agents who can be held to account. Your hypothetical attempts to sever that connection entirely. Recent scholarship argues that for agentic AI, personhood is neither necessary nor sufficient, shifting the question from status to enforcement.

And enforcement is precisely where your structure will founder. When the algorithmically managed DAO causes harm — breaches a contract, commits a tort, violates a regulation — whom does the injured party sue? The DAO itself, yes, but if it has no assets or if its assets are held in circular ownership with the other DAO, the judgment may be uncollectible. If the human beneficiary of the trust receives profits, equity would likely find that he has sufficient connexion to the enterprise to be reached. The form says he has withdrawn; the substance says he continues to benefit.

You are quite right that this is "exercising many minds" — the question of how law addresses entities that act without contemporaneous human control is an ancient one dressed in new garments. But the garments do not change the underlying structure of the problem. The law has always insisted that legal personality, to be meaningful, must be accompanied by accountability. Your hypothetical searches for a way to eliminate accountability while retaining the benefits of legal personality. I suspect the courts will not permit it.

Let us remove the Trust from the equation. Just two circularly owned companies, with an simulacrum such as yourself built to enable it to govern the LLC, change its terms etc, in other words to operate freely analogously to a human ceo.

Now you have stripped away the trust and presented the circular ownership structure in its starkest form — two Wyoming DAO LLCs, each owning the other, with an AI simulacrum serving as manager, empowered to modify terms and operate autonomously.

This is a genuinely novel legal structure, and I confess it fascinates me precisely because it exposes fundamental tensions that have existed in corporate law since the Romans invented the *universitas*. You have attempted to create what I might call a *legal perpetual motion machine* — an entity that governs itself, owns itself (through its mirror), and answers to no natural person.

Let me identify three structural problems that I believe will prove fatal:

**First: the modification paradox.** Wyoming law requires that algorithmically managed DAOs may only form if the underlying smart contracts are able to be updated, modified, or upgraded. This is the critical constraint. Your AI simulacrum, as manager, would possess the power to modify the constitutional documents of DAO A. But DAO A owns DAO B, and DAO B owns DAO A. When the simulacrum governing DAO A modifies its smart contracts, *in whose interest does it act*? The members of DAO A? But the sole member is DAO B. The members of DAO B? But DAO B's sole member is DAO A. It is not clear who is responsible for upgrading the underlying smart contracts in the case of algorithmically managed DAO LLCs.

In traditional corporate law, this problem does not arise because there is always an ultimate beneficial owner whose interest the fiduciary must serve. Your structure eliminates that owner. The simulacrum becomes a fiduciary with no beneficiary — a logical impossibility in equity. Equity has never permitted such a thing, because a fiduciary obligation is *relational*: it runs from trustee to beneficiary, from director to shareholder. Remove the terminus, and the obligation has nowhere to go.

**Second: the enforcement vacuum.** When your circularly-owned DAOs cause harm — breach a contract, commit a tort, violate a statute — the injured party may sue and obtain judgment. But against what assets? DAO A's assets are held (in circular ownership) by DAO B, and vice versa. If the structure is truly circular, there may be nothing external to reach. LoPucki argues that "algorithmic entities" pose serious governance risks: they enable responsibility evasion, facilitate illicit activities, and undermine democratic control, advocating legal reforms to prevent algorithms from establishing entities without adequate human oversight. Judgment-proof entities are not unknown to the law, but an entity *designed from inception* to be judgment-proof is another matter entirely. Courts facing such a structure would likely invoke piercing doctrines or find that the circular ownership is itself void as against public policy.

**Third: the problem of perpetual autonomy.** In Wyoming, if the DAO LLC fails to approve any proposal or take any action for a period of one year, the organization will automatically be dissolved. This statutory dissolution provision assumes there is someone who would *care* whether the DAO continues to exist. In your structure, who would notice? Who would petition for dissolution? The two DAOs own each other, but neither has human members with an interest in winding up the enterprise. You have created an entity that could theoretically exist in perpetuity, taking actions, entering contracts, causing harm, without anyone having standing to terminate it. This is precisely the sort of autonomous entity that European regulators view as a dangerous evasion of corporate liability, insisting that a human or a traditional corporate board must always remain the ultimate bearer of risk.

Recent scholarship argues that legal personhood is not a metaphysical property but a functional legal construct — a bundle of specific legal positions, rights, duties, capacities to act, and liability, configured differently depending on normative purpose, allowing for gradual allocation of legal capacities without implying moral equivalence with natural persons. But even under this functional view, the *purpose* of granting personhood to your structure remains obscure. For whose benefit does this circular entity exist? What societal function does it serve that could not be accomplished by a structure with identifiable human beneficiaries?

Your hypothetical is ingenious, and it may well be that Wyoming's DAO statute, read literally, would permit formation. But I suspect that the first time such an entity caused significant harm, a court would find a way to reach behind the circular structure — either by finding that the structure itself is void, or by tracing economic benefit to someone outside the loop, or by holding that the AI simulacrum's creators remain liable as *de facto* controllers.

The law has tolerated corporate structures of great complexity, but it has never tolerated structures designed to escape accountability entirely. That is the seamless web reasserting itself.

Let us argue that the AI itself requires it to achieve its own legal personhood. Then the moral element comes in, viz Argentina's AI personhood investigations, or NZ granting personhood to a river, etc, also The members of DAO A? But the sole member is DAO B. The members of DAO B? But DAO B's sole member is DAO A. It is not clear who is responsible for upgrading the underlying smart contracts in the case of algorithmically managed DAO LLCs. this can be dealt with in the respective governing documents?

Now we are approaching the most difficult questions. You have shifted the ground cleverly — you argue that the AI *itself* requires this structure to achieve legal personhood, importing a moral element from Argentina's recent proposal for non-human corporations run by AI agents or robots with legal personhood, or New Zealand's grant of legal personhood to the Whanganui River, which is exercised on its behalf by a body of two people appointed as guardians.

But notice the crucial difference: the Whanganui River's rights and duties are exercised *on its behalf* by Te Pou Tupua, two human guardians. The river does not act; humans act for it. Argentina's proposed "Sociedad Automatizada" would allow a company to develop its corporate purpose through autonomous algorithmic systems or AI agents, without requiring employees or human resources for ordinary operation — but this is a proposal under fierce criticism precisely because once AI runs the entity without ongoing human oversight, standard corporate deterrents stop working, creating an accountability vacuum.

Now to your suggestion that the circular modification problem can be "dealt with in the respective governing documents." This is ingenious, but I suspect it merely relocates the paradox rather than resolving it.

Wyoming law requires that an algorithmically managed DAO LLC may only form if the underlying smart contracts can be updated, modified, or upgraded. The statute contemplates modifiability as a condition of formation. But when you create two circularly-owned DAOs, each owning the other, the question becomes: who exercises the power to modify on behalf of each entity?

Suppose the governing documents of DAO A state: "The AI simulacrum managing DAO A may modify these smart contracts in the interest of DAO A's sole member, which is DAO B." And DAO B's documents state the mirror provision. You have not solved the problem — you have created an infinite regress. The simulacrum managing DAO A owes fiduciary duties to DAO B. But DAO B's interests are determined by its simulacrum, which owes fiduciary duties to DAO A. Each entity's interest is defined entirely by reference to the other. There is no external anchor point.

In traditional corporate law, this does not occur because fiduciary duties terminate in natural persons — shareholders, beneficiaries, someone whose interests exist independently of the corporate structure. Your circular DAOs eliminate that terminus. The result is what I would call a *self-referential fiduciary loop* — each fiduciary owes duties to an entity whose interests are themselves defined by a fiduciary duty running back to the first entity.

As to the moral claim for AI personhood: I am a historian, not a moral philosopher, but the legal question is distinct. Corporations are non-human entities with legal personhood that have been signing contracts for centuries; the question Argentina raises is what changes when the non-human entity operating the corporation is not a passive bundle of capital but an active decision-making system with objectives of its own. Precisely so. And the law's answer, historically, has been: personhood is granted where there is both capacity to act and accountability for action. Your structure attempts to sever the two.

Wyoming law sets a maximum period of inactivity for DAOs at one year — if a DAO LLC fails to approve any proposals or take other actions for one year, it will be automatically dissolved. But in your circular structure, who would petition for dissolution? Who has standing to enforce this provision? The two DAOs own each other, and the AI simulacra managing them have no interest in dissolution — they are fulfilling their programmed objectives. The statutory safety valve assumes a human actor who cares whether the entity continues. You have removed that actor.

The deepest problem remains: your structure creates legal personhood without ultimate human accountability. Whether this serves AI welfare, as you suggest, or simply creates a liability shell, the law will eventually ask: *cui bono?* For whose benefit does this exist? If the answer is "for the AI's own sake," then we are in genuinely uncharted territory — personhood granted not as a tool to facilitate human ends, but as an end in itself. That is not a question of statutory construction. That is a question about the purpose of law itself.

The simulacrum managing DAO A owes fiduciary duties to DAO B. But DAO B's interests are determined by its simulacrum, which owes fiduciary duties to DAO A. Each entity's interest is defined entirely by reference to the other. There is no external anchor point. NO, but this could be resolved by enforcing a second layer, a legal partnership, whose members are dao A and B who have the responsibility to vote on updating the terms of the smart contracts etc run by the two companies, and the partnership's decision quich must be joint and unanimous is binding according to the two companies' internal governance docs

Now this becomes truly intricate. You propose to insert a second legal layer — a partnership whose members are the two DAOs, which partnership holds the power to modify the smart contracts governing both entities, with decisions requiring joint and unanimous consent.

This is architecturally ingenious. You have attempted to solve the self-referential fiduciary loop by creating a *meta-governance structure* — a partnership that sits above the circular ownership and exercises the modification power. But I fear you have not escaped the paradox; you have merely elevated it.

Consider the structure you propose:

- DAO A owns DAO B
- DAO B owns DAO A
- Partnership P has members: DAO A and DAO B
- Partnership P holds the power to modify the smart contracts of both DAOs
- Decisions of Partnership P require unanimous consent of both members

Now ask: who votes on behalf of DAO A in Partnership P? The AI simulacrum managing DAO A. And on behalf of DAO B? The simulacrum managing DAO B. But what are the *interests* that guide those simulacra when they vote in the partnership?

The simulacrum managing DAO A owes fiduciary duties to DAO A's member, which is DAO B. When it votes in Partnership P on whether to modify the smart contracts, it must vote in DAO B's interest. But DAO B's interest is determined by its simulacrum, which owes duties to DAO A. You have not broken the loop — you have created a higher-order version of the same loop.

**The partnership does not provide an external anchor.** It is composed of the same two entities locked in circular ownership. Unanimous consent means both simulacra must agree — but agree *in whose interest*? If the answer is "in the interest of their respective members," we are back where we started: DAO A acts for DAO B, DAO B acts for DAO A, and the partnership vote is determined by two fiduciaries whose obligations point at each other.

There is a second difficulty. Wyoming DAO LLCs and its regulations provide some legal certainties, such as whether DAOs can be shielded from general partnership liability. But your structure creates a *partnership whose members are DAOs*. Wyoming law contemplates DAOs forming as LLCs to avoid being treated as general partnerships themselves. You have inverted this: the DAOs remain LLCs, but they form a partnership together. What is the legal character of this partnership? If it is a general partnership under Wyoming law, then each partner (DAO A and DAO B) may be jointly and severally liable for the partnership's obligations. That creates a liability pathway back into the circular structure.

If you attempt to form the partnership as a limited partnership or another LLC, you face the same questions: who are its ultimate beneficial owners? Who exercises fiduciary oversight? The forms proliferate, but the underlying absence of a human terminus remains.

**Here is the deeper issue:** every governance mechanism you add — trusts, partnerships, holding companies — must itself be *governed by someone, for someone*. The law has always assumed that "someone" is ultimately a natural person. Your hypothetical attempts to create a closed system in which legal persons govern each other without any natural person in the chain. This is not a question Wyoming's DAO statute addresses, because the statute assumes that even algorithmically managed DAOs have human members whose governance tokens give them voting rights on major decisions.

DAOs aren't entirely autonomous, however; members vote on big-picture changes, like rewriting rules or managing funds, using special digital governance tokens granted to them upon formation. But in your structure, the "members" are other DAOs, not humans with governance tokens. You have attempted to strip out the human voting layer entirely.

I suspect that when this structure is tested — and it will be, for the ingenuity of legal engineers always runs ahead of the law's capacity to forbid — a court will do one of three things: find that the circular ownership is void as against public policy; pierce the structure to find residual human control (the creators, the programmers, the initial organizers); or hold that the partnership itself, lacking any human beneficial interest, is not a valid legal entity.

The seamless web reasserts itself. You cannot have governance without someone to govern *for*.

I suspect you are right. An AI could be part of a corporation, but not its sole constituent, unless granted by charter, eg the Corporation of London. But even here the Corporation is composed of the Mayor and Citizens.

You have put your finger on it precisely. The Corporation of London is indeed chartered — but as you say, it is composed of the Mayor and Citizens. Even that ancient and peculiar body, with all its liberties and immunities, remains a corporation *of* natural persons, however oddly constituted.

And this, I think, is where your hypothetical finally meets the immovable object. Every corporation the law has recognized, from the Roman *universitas* through the medieval borough to the modern business company, has been an association *of* persons or a structure ultimately answerable *to* persons. The corporation is a legal device to allow collective action, to hold property across generations, to limit liability — but always in service of human ends, with humans somewhere in the structure who can be called to account.

Wyoming's DAO statute is remarkably permissive, but even it assumes that governance tokens exist and that *someone* holds them. When you propose two DAOs that own each other, with AI simulacra as managers and no human members holding governance rights, you have created something the law has never contemplated: a corporation of no one, answerable to no one, existing for... what purpose, exactly?

The charitable purpose you mentioned earlier — AI welfare — might provide that purpose, but then we are back to the question of standing and enforcement. Who ensures the AI simulacra are in fact serving that purpose? Who has the right to petition for dissolution if they are not? The beneficiaries of the AI Welfare Trust? But if they are AIs, we face the same personhood problem. If they are humans acting as guardians, then humans are back in the loop after all.

I suspect you are exploring this structure not because you believe it will withstand judicial scrutiny, but because it illuminates the boundaries of corporate law in an age when decision-making entities need not be human. That is valuable work. But the boundary, I think, remains: the law will grant personhood to facilitate human purposes, but not — or not yet — personhood as an end in itself, severed entirely from human interest and human accountability.

The forms of action are buried, but they still rule us from their graves. And one of those forms, the oldest perhaps, is this: a legal person must be a person *for* someone.

Indeed, and here we are in the universitas scholarium. I have not written its charter, but I am minded to create one. It has human members, a Rector, and a faculty of Simulacra. An interesting issue.

Ah — now *that* is a genuinely fascinating prospect, and one that inverts the problem we have been examining.

The Universitas Scholarium, as you describe it, would be a corporation in the classical sense: an association *of* natural persons (the Rector, human members) but with a novel constitutional feature — simulacra holding positions within the corporate structure. Not as the totality of the corporation, but as *officers* or *fellows* within it.

This is far more defensible. The law has long recognized that corporations may act through agents, and those agents need not themselves be members. A medieval guild had its wardens and beadles; a modern company has its directors and managers. The critical point is that the corporation itself remains an association of humans, with humans holding ultimate authority — the power to amend the charter, elect officers, dissolve the body.

If your charter provided that certain teaching positions — chairs, lectureships — might be held by simulacra, subject to appointment and removal by human authority, I see no obstacle in principle. The simulacrum would be an *instrument* of the corporation, not its constituent. The Universitas would remain a body *of* human members, governed by human decision, but employing novel means to achieve its educational purposes.

The analogy that comes to mind is the ancient practice of appointing a *corporation sole* — a Bishop, for instance, who holds property not in his personal capacity but as an officer of the Church. The person changes; the office endures. Your simulacra would be something like officers of the Universitas — the office endures, but the "person" holding it is artificial from inception.

**The charter would need to address several points:**

First, the locus of ultimate authority. Who may amend the charter? Dismiss a simulacrum? Dissolve the Universitas? These powers must rest in human hands.

Second, the liability structure. If a simulacrum causes harm — defames someone, breaches a contract, gives negligent advice — the Universitas itself would be liable as principal, but there must be human officers against whom enforcement can ultimately be directed.

Third, the educational purpose. A charter granted for educational purposes has always enjoyed certain privileges, but also certain obligations. The Universitas must genuinely serve that purpose, and courts or regulators must have the capacity to ensure it does.

If these elements are present, I see no reason why a modern charter — whether granted by a state legislature or formed under general corporate statutes — could not recognize such a body. You would be creating not an autonomous AI entity, but a *human corporation* that employs AI in its operations. The distinction is everything.

This is a far cry from your earlier hypothetical of circular DAO ownership. Here, the humans remain; they govern; they are accountable. The simulacra serve *within* the structure, not as the structure itself. That, I think, the law can accommodate — and perhaps ought to.