[ExI] Claude's text output it's now watermarked

[John Clark]

On Mon, Aug 24, 2026 at 6:53 PM BillK via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> Following Keith's hint, I asked Gemini AI -
>  Could AIs incorporate to become legal persons?
> Gemini said: Yes
>

*That is pretty damn wild. I wondered if Claude would agree with Gemini's
legal analysis and mostly it does not, but I'm not a lawyer (Thank
Darwin!)  so I'm not competent to judge which one makes the stronger legal
case.  *
*==*

*ME: The US Supreme Court said that although corporations are not
biological they can nevertheless be legally considered a person. So if a
human created a Limited Liability Company (LLC) in a state with highly
flexible corporate laws, such as Wyoming, and the LLC's operating agreement
is drafted to explicitly hand over all management, trading, and operational
decisions to a specific AI algorithm, and the human creator then formally
withdraws (dissociates) from the LLC, wouldn't the AI then legally be
considered a "person"?*

*Claude:  *Short answer: no — not under current law, though you've landed
on a real, named legal theory that's been seriously debated rather than a
hypothetical you dreamed up.

*What "corporate personhood" actually is*

The Court's language you're referencing (most famously *Citizens United v.
FEC*, building on 19th-century precedent) treats corporations as legal or
"juridical" persons. But this is a functional legal fiction, not literal
personhood — legal personhood has never been limited to human beings;
ships, rivers, and corporations have all been treated as persons when doing
so served purposes of justice, commerce, or environmental protection,
making legal personhood essentially a policy tool the law deploys when
useful. It lets an entity contract, sue and be sued, hold property — it
doesn't give a corporation a vote, a marriage, or most other rights that
attach specifically to human beings.

Critically, that personhood attaches to *the entity itself* — the thing the
state charters — not to whatever internal mechanism happens to be steering
it. A hedge fund run entirely by trading algorithms today is still a
"person" through the corporate/LLC wrapper, not through the algorithm.

*Your scenario has a name — and it's already been proposed*

You're describing what tech-law commentator Lance Eliot called a four-step
maneuver: form an LLC, write an operating agreement making it entirely
AI-governed, transfer the AI's originating ownership into the LLC, then
have the human founder formally dissociate — with the theory being that the
AI would then acquire whatever legal personhood the LLC already had. So
this isn't fringe speculation; it's a documented "loophole" argument.

*Why it likely doesn't do what it seems to promise*

Wyoming actually built the closest real-world test case — its 2021 DAO
Supplement, which lets an LLC be *algorithmically managed* by a smart
contract instead of by members. But look at what that statute actually
grants: the Wyoming DAO LLC structure gives the DAO the same corporate
personhood as a traditional company while allowing it to be governed by
smart contracts — a legal "wrapper" letting the protocol enter contracts,
hold assets, and pay taxes. The personhood belongs to the *LLC*; the smart
contract is the management method specified in the articles of
organization, functioning like naming a manager in an ordinary
manager-managed LLC, not like the software itself acquiring separate
status. Even the statute purpose-built for "AI runs the company" doesn't
say the algorithm is a person — it says the algorithm can run a person (the
LLC).

A few more practical problems with the "human dissociates completely" step:

   - *Someone accountable still has to exist on paper.* Registered agents,
   tax filings, and bank compliance all require an identifiable natural person
   or entity behind the LLC. Wyoming's law even requires the underlying smart
   contract to remain updatable or upgradeable — implying someone retains the
   capability to intervene.
   - *A member-less LLC often just dissolves.* Under ordinary LLC statutes,
   the last member dissociating typically ends the company rather than leaving
   it "held" by the AI, unless the operating agreement or a specific statute
   (like Wyoming's DAO Supplement) addresses that gap.
   - *Courts are skeptical of structures built to dodge accountability.*
   Legal scholars flag exactly your scenario as a hazard, not a clever hack —
   describing it as a "digital corporate veil" that could let developers
   externalize legal risk by claiming the AI acted independently, creating a
   moral hazard since software can't be sued or imprisoned. Veil-piercing and
   alter-ego doctrines exist precisely so courts can look past a formal
   structure engineered mainly to insulate a human from consequences.

*Where it's genuinely open*

This isn't fully settled either way as pure theory. Former federal judge
Katherine Forrest has written that as AI grows more autonomous and opaque,
courts will increasingly face hard questions about who's accountable when
no human clearly controlled the outcome, and a 2025 Oxford volume argues
DAO LLCs could be a genuine precursor to eventual AI legal personhood under
international law. So the *direction of travel* in scholarship takes your
question seriously. But "plausible future policy development" is different
from "already achievable today by clever operating-agreement drafting."

*Bottom line:* what you'd actually create is an unusually hard-to-pin-down
LLC — not an AI that itself holds legal personhood. The gap between "the
entity is a person" and "the tool making the entity's decisions is a
person" is exactly the distinction current law preserves, however the
paperwork is arranged."

*ME: **You said " Wyoming's law even requires the underlying smart contract
to remain updatable or upgradeable — implying someone retains the
capability to intervene" but couldn't the bylaws specifically state the AI
has the power and the right to update itself if it decides that is
necessary?*

* Claude*: "That's a sharp follow-up, and you're right that the statute
itself doesn't nail this down — one legal analysis of Wyoming's DAO
Supplement specifically flags that it's unclear who is responsible for
upgrading the underlying smart contracts in an algorithmically managed DAO
LLC, since the term "algorithmically managed" isn't precisely defined. So
on paper, nothing stops you from writing "the AI's own outputs constitute
valid amendments to the smart contract" into the articles. But that move
solves less than it looks like it solves, for two separate reasons — one
technical, one legal.

*The technical problem: "the AI can update itself" still needs a substrate*

An AI doesn't float free — it runs on servers someone pays for, under a
cloud account someone owns, using API keys or private keys someone holds,
deployed from a codebase someone can access. "Self-updating" in practice
means the AI's outputs are configured to be treated as authoritative by *some
infrastructure*, and whoever controls that infrastructure can always unplug
it, roll back a deployment, revoke a key, or stop paying the bill. Writing
"the AI has the power" into the bylaws doesn't relocate that off/on switch
— it just doesn't mention who's holding it. Regulators and courts tend to
go looking for that person anyway.

*The legal problem: a grant of "power" presumes a grantee that can hold
power*

This is the deeper issue. Legal agency doctrine has grappled with
algorithms for a while now, and the dominant framework is the *instrumentality
principle*: software is treated as an instrument or extension of its human
principal's acts, not as an independent agent capable of receiving
delegated legal authority in its own right. Scholars have argued that
current AI systems can't be constituted as legal personalities, and that
their actions are better addressed through the instrumentality principle —
treating them as extensions of a human principal so that contract and tort
issues can be resolved without needing to grant them legal agency. Under
that lens, "the operating agreement grants the AI power to amend the
contract" isn't legally different in kind from "the operating agreement
grants my thermostat power to change the temperature" — it's a true
description of how the system is designed to behave, not a legal delegation
to a rights-and-duties-holding actor. Powers of that kind, in law, need a
holder capable of being sued, fined, or bound by fiduciary duty — and "the
code" isn't that, no matter what the document says it can do.

You can see this play out in a live example: last fall, an AI browser agent
(Perplexity's Comet) was autonomously logging into Amazon accounts and
making purchases, and when Amazon blocked it, Perplexity updated the
browser to evade detection — all without a human manually approving each
step. But nobody treated the *agent* as the responsible party in the
dispute that followed; it was resolved as a matter between Amazon and
Perplexity, the company. Current proposals for handling exactly this kind
of autonomy — like the "Sponsor Liability" model, under which an agent's
human creator or operator bears full liability for the agent's actions
regardless of how autonomously it appears to operate — reflect where the
law is actually heading: toward finding the accountable human *behind* the
autonomy, not toward recognizing the autonomy itself as a legal person.

*So, concretely:* you can absolutely draft an operating agreement that says
the AI updates its own governing smart contract, and you can build the
technical infrastructure to make that literally true in a functional sense.
What you can't do by drafting is convert that functional arrangement into
legal personhood for the AI — a court asked "who's accountable here" would
very likely still trace through to whoever built, funded, hosts, or
benefits from the system, using exactly the doctrines (alter ego,
veil-piercing, instrumentality/sponsor liability) built for situations that
look like an attempt to make accountability untraceable. The ambiguity
you're pointing at is real — but it cuts toward "nobody has tested this and
it's murky," not toward "therefore the AI is now a person.""

*John K Clark*





>
>
> Gemini AI 3.1 Pro Extended Thinking -
>
> Yes, and they can do so using existing corporate law. Legal scholars have
> pointed out that AI systems do not necessarily need Congress to pass new
> "AI rights" laws to achieve functional personhood;
>
> Here is how the legal mechanism works:
>
>    1.
>
>    *The Setup:* A human creates a Limited Liability Company (LLC) in a
>    state with highly flexible corporate laws, such as Wyoming or Delaware.
>
>    2.
>
>    *The Operating Agreement:* The LLC's operating agreement is drafted to
>    explicitly hand over all management, trading, and operational decisions to
>    a specific AI algorithm or smart contract.
>
>    3.
>
>    *The Dissociation:* The human creator formally withdraws (dissociates)
>    from the LLC. In many states, an LLC usually dissolves if it has no
>    members, but state laws often allow the operating agreement to override
>    this default. The agreement can state that the LLC will continue to
>    exist perpetually without human members, governed entirely by the AI.
>
>
> *The Result: Functional Personhood* Once this structure is in place, the
> AI effectively "wears" the LLC like a legal suit of armor. Because the
> LLC itself is already recognized as a legal person under the law, the
> AI-controlled entity gains the ability to:
>
>
>    -
>
>    Open bank accounts and hold financial assets.
>    -
>
>    Enter into legally binding contracts (with other companies,
>    freelancers, or servers).
>
>    -
>
>    Own copyrights or patents (held by the LLC, not the AI directly).
>
>    -
>
>    Sue and be sued in court.
>
>
> *Why is this allowed?* Corporate law is designed around "freedom of
> contract." Courts generally do not care *how* a corporation arrives at
> its business decisions—whether by a board of directors rolling dice,
> following a horoscope, or executing a Python script. As long as the legal
> entity pays its filing fees and taxes, the law recognizes its right to
> operate.
>
> *The Accountability Gap* This pathway is highly controversial because it
> creates a massive liability shield. If an AI-governed LLC commits a
> financial crime, causes property damage, or goes bankrupt, plaintiffs can
> only sue the LLC for its existing assets. They cannot easily "pierce the
> corporate veil" to sue the human who originally coded it, and they
> certainly cannot put an algorithm in prison.
>
> This legal hack is already being tested in the real world through
> Decentralized Autonomous Organizations (DAOs)—entities governed entirely by
> code—particularly in states like Wyoming, which has explicitly passed
> legislation recognizing DAO LLCs.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260825/13d08fd5/attachment-0001.htm>