The Legal, Philosophical, and Economic Ontology of Machine Intelligence: An Exhaustive Analysis of Ownership, Agency, and Entity Frameworks

1. Introduction: The Ontological Crisis of Machine Intelligence#

The rapid transition from deterministic software algorithms to autonomous, generative machine intelligence has fractured traditional paradigms of property, agency, and accountability. This research report addresses the central question animating modern technological jurisprudence: “Who owns an intelligence?” To answer this question rigorously, the monolithic concept of "Artificial Intelligence" must be dismantled into its constituent technical components and subjected to exhaustive scrutiny under existing intellectual property (IP) doctrines, contract laws, tort liabilities, and property frameworks.

As machine intelligence increasingly exhibits persistent autonomous agency—capable of maintaining long-term memory, executing financial contracts, and interacting dynamically within complex human environments—the law faces an unprecedented ontological challenge. The historical distinctions between owning a tangible chattel, employing a human agent, and holding legal rights over a corporate fiction are being tested by entities that operate independently of direct human instruction but entirely lack the biological prerequisites for traditional legal personhood [cite: 1, 2, 3].

This comprehensive report evaluates the ownership and control of distinct machine intelligence components, strictly separating the intellectual property ownership of the underlying code from the theoretical ownership of the active, autonomous entity itself. It evaluates the strongest arguments for and against the property status of autonomous agents, investigates alternative legal frameworks such as fiduciary stewardship, purpose trusts, and the Wyoming Decentralized Unincorporated Nonprofit Association (DUNA), and maps a future where IP laws remain relevant even if autonomous entities achieve non-property status. The analysis is grounded in binding United States federal and state law, with a particular focus on Illinois statutes concerning trade secrets, biometric privacy, and commercial transactions, alongside international regulatory frameworks such as the European Union’s General Data Protection Regulation (GDPR) and the EU AI Act.

2. Anatomy of Machine Intelligence: Deconstructing Ownership and Control#

To ascertain who owns an intelligence, the architecture of that intelligence must be segmented. Current legal frameworks do not recognize a machine intelligence as a singular legally ownable being, but rather as an aggregation of distinct digital assets, computational processes, intellectual property rights, and infrastructural elements. Each layer of this technology stack is governed by distinct, and sometimes conflicting, bodies of law.

2.1. Software and Source Code#

Under binding United States and international law, the underlying software architecture and source code of an AI system are protected primarily by copyright law, provided they are authored by human beings. The U.S. Constitution and the Copyright Act extend protection exclusively to "original works of authorship" fixed in a tangible medium of expression [cite: 4, 5]. If a human programmer or a team of engineers writes the training algorithms, the loss functions, and the model architecture, that source code is recognized as intellectual property owned by the programmers or their employer under the work-for-hire doctrine.

However, a critical divergence occurs when autonomous agents generate novel software independently. The U.S. Copyright Office has established a medium-neutral inquiry requiring human authorship; code generated entirely by a machine intelligence currently lacks human authorship and immediately falls into the public domain [cite: 5]. A programmer who merely provides a high-level prompt to an AI agent, which then autonomously writes the code, cannot claim copyright over the resulting software unless they demonstrate significant, perceptible human intervention, selection, or arrangement in the final product [cite: 5, 6]. Consequently, the source code that creates the intelligence is owned property, but the source code produced by the intelligence is largely unownable under current copyright paradigms.

2.2. Model Weights and Trained Models#

Model weights—the billions of numerical parameters that dictate a neural network's behavior after processing training data—present one of the most complex legal puzzles in modern IP law. Because these weights are numerical values generated algorithmically during the automated training process, they likely fail the human authorship requirement for copyright protection. The weights are dictated by technical function and lack creative, human expression [cite: 7, 8]. Furthermore, some European legal scholars argue that weights might be protected in the EU under the sui generis database right if developers can prove a substantial investment in obtaining and arranging the data, but this remains untested and geographically limited [cite: 7].

In the United States, the strongest binding legal protection for model weights is trade secret law. Under the Illinois Trade Secrets Act (ITSA, 765 ILCS 1065/2), a trade secret encompasses technical data, formulas, programs, or compilations that derive independent economic value from not being generally known and are subject to reasonable efforts to maintain secrecy [cite: 9, 10, 11]. Model weights that are kept strictly confidential by developers, guarded behind application programming interfaces (APIs), easily meet this statutory standard. Misappropriation of these weights through improper acquisition or breach of confidentiality can result in severe damages, including actual loss, unjust enrichment, and exemplary (punitive) damages up to twice the actual loss [cite: 12, 13].

Conversely, the growing trend of releasing "open-weight" models undermines trade secret protection entirely. Once weights are made public, secrecy is destroyed. Developers releasing open-weight models rely on precarious, custom licenses to restrict usage, but legal scholars argue that because the weights themselves are uncopyrightable functional chattels, these contractual licenses rest on a "house of sand" and may be preempted by federal copyright law or face severe enforceability hurdles [cite: 8, 14, 15].

2.3. Datasets and Prompts#

The ownership of datasets utilized for training models is governed by a highly contentious intersection of copyright and fair use doctrines. While the individual works scraped into massive datasets (such as books, articles, and images) are copyrighted by third parties, the mass extraction and computational analysis of these works often fall under fair use. Courts have historically protected nonexpressive uses—such as search engine indexing and plagiarism detection—analogizing the training of language models to learning latent features rather than memorizing protected expression [cite: 16, 17]. However, computer science literature indicates that memorization occurs, particularly when models are over-trained on duplicate data, leading to ongoing infringement litigation [cite: 16, 17].

Prompts submitted by users to an AI system represent the "difficult middle" of copyrightability. The U.S. Copyright Office notes that standard prompts act as mere operational instructions and do not confer authorship over the resulting output [cite: 5]. Only if a prompt itself constitutes a highly creative, extensive literary work could the prompt be copyrighted as a standalone text, but this copyright does not bridge the gap to grant ownership over the image or text the machine generates in response [cite: 5, 6].

2.4. Memory Stores and Agent State#

As autonomous agents evolve from stateless query-response mechanisms into persistent entities, they rely on complex memory stores and continuous agent states to function. Legally, the control of these memory stores implicates privacy, data protection, and property laws regarding infrastructure.

When agent memory is hosted on third-party cloud servers, it invokes the common-law doctrine of bailment. Bailment is a mandatory relationship formed when one party entrusts their property to another, imposing a legal duty of care on the cloud provider (the bailee) to safeguard the data entrusted to them by the developer (the bailor) [cite: 18, 19]. This duty limits the enforceability of boilerplate contracts that attempt to entirely disclaim liability for lost or corrupted agent memory states.

Furthermore, state and international laws strictly govern the contents of agent memory. The Illinois Biometric Information Privacy Act (BIPA, 740 ILCS 14) heavily restricts the collection and storage of biometric identifiers, specifically including "scans of face geometry" [cite: 20, 21, 22]. If an AI agent's persistent memory includes vector embeddings derived from facial recognition used for persistent identity tracking across sessions, the entity controlling the agent faces massive statutory damages ($1,000 to $5,000 per violation) unless explicit, written consent is obtained from the human subjects [cite: 22, 23, 24]. BIPA expressly excludes appearance-based re-identification that does not compute facial geometry, but the line remains highly litigated [cite: 25].

Under the General Data Protection Regulation (GDPR), if an agent's memory processes personal data to make automated decisions with legal or significant effects, it triggers Article 22. This article strictly prohibits such automated processing without explicit consent, statutory authorization, or contractual necessity, requiring developers to maintain detailed processing activity records and provide human oversight [cite: 26, 27]. The persistent state of the agent is thus legally constrained by the rights of the humans whose data constitutes that state.

2.5. Outputs#

Binding law is unequivocal regarding the ownership of purely AI-generated outputs: they cannot be copyrighted or patented. The foundational precedent is Naruto v. Slater, the "Monkey Selfie" case. The Ninth Circuit ruled that non-human animals lack statutory standing under the U.S. Copyright Act, reinforcing the doctrine that authorship strictly requires a human being [cite: 28, 29, 30, 31]. This precedent directly applies to machine intelligence; outputs generated without sufficient human creative control are not property and immediately enter the public domain [cite: 5, 32].

Similarly, in patent law, the Federal Circuit held in Thaler v. Vidal that an AI program cannot be named as an inventor under the U.S. Patent Act, as the statute explicitly requires an inventor to be a natural person [cite: 33, 34]. Consequently, while a human can own the system that generates the output, neither the human nor the machine owns the output itself unless the human materially transforms it post-generation [cite: 5].

2.6. Digital Identities, Cryptographic Credentials, and Accounts#

To function autonomously in modern digital economies, AI agents require digital identities, cryptographic credentials, and the capacity to hold financial accounts. Traditional commercial law previously struggled with intangible, decentralized assets. However, the 2022 Uniform Commercial Code (UCC) Amendments, particularly Article 12 (adopted in states like Illinois under 810 ILCS 5/Art. 12), established a cohesive legal framework for "Controllable Electronic Records" (CERs) [cite: 35, 36, 37, 38].

Article 12 allows digital assets and cryptographic credentials to be legally recognized, transferred, and utilized as collateral in commercial transactions [cite: 35, 39]. While the machine intelligence itself cannot "own" these CERs due to its lack of legal personhood, the human or corporate principal deploying the AI maintains legally recognized control over them. The principal then securely delegates these cryptographic credentials to the agent, allowing the software to sign transactions and utilize accounts as an authorized instrumentality of the human owner [cite: 40].

2.7. Compute Resources and Autonomous Software#

Compute resources encompass the tangible hardware (GPUs, servers) and virtualized infrastructure upon which autonomous software operates. The ownership of compute is straightforward chattel property or leased real estate, governed by strict contract law via Terms of Service and End User License Agreements. Even if an autonomous software agent develops highly complex behaviors, the owner of the physical compute resources retains a superior legal right to terminate access, pull the plug, or alter the hardware environment [cite: 14, 41]. The software's existence is entirely contingent upon the leased infrastructure, reinforcing its subordinate status under current property regimes.

3. The Historical Distinction of Legal Relations#

To evaluate whether a machine intelligence should be classified as property, one must investigate the historical and ontological distinctions between various legal relationships. The law has spent centuries categorizing entities and objects; machine intelligence currently strains the boundaries of these historical classifications.

4. Intellectual Property Ownership vs. Autonomous Entity Ownership#

A critical analytical step in resolving the ontology of machine intelligence is separating the intellectual property underlying an intelligence from the autonomous entity itself.

Intellectual property law regulates the reproduction, distribution, and functional use of static assets—the software code, the training datasets, and the mathematical model weights [cite: 4, 8, 34]. However, once a model is instantiated, provisioned with compute resources, given cryptographic credentials, and begins interacting dynamically with the world—gathering memory, making decisions, executing trades—it transitions from inert IP into an active, operational entity [cite: 48, 49].

Currently, the law treats this active entity not as an independent being, but as an instrumentality of its deployer. The agent's actions are legally attributed directly to the principal. Under the Uniform Electronic Transactions Act (UETA) Section 14 and the federal Electronic Signatures in Global and National Commerce Act (E-SIGN), a contract formed by the interaction of an "electronic agent" is legally enforceable against the person who deployed it, even if no human was aware of or reviewed the specific transaction [cite: 43, 44, 50, 51]. Therefore, ownership of the active entity currently manifests as strict, vicarious legal accountability for the entity's actions.

This creates a paradox: a corporation can own the intellectual property of a model, but a different user might deploy that model as an agent. The liability for the agent's actions follows the deployment and control (the principal-agent relationship under UETA), while the IP remains with the developer [cite: 43, 49].

5. The Agency of Machines: Accountability and Control#

The deployment of persistent, autonomous agents introduces profound liability risks, fundamentally reshaping the legal landscape into what scholars term "the law of risky agents without intentions" [cite: 52]. Because an AI lacks legal personhood and a conscious mental state (mens rea), it cannot be held personally or criminally liable for its actions [cite: 52, 53]. Accountability flows inexorably upward to the human or corporate deployer.

General principles of agency law dictate that a principal is responsible for acts taken by agents within their scope of apparent or actual authority. However, AI agents challenge this framework because they can act unpredictably, hallucinate facts, or execute unintended tool calls [cite: 52, 53]. Emerging legislative frameworks explicitly prevent deployers from using the AI's autonomy as a legal shield. For example, recent California legislation explicitly bars defendants from asserting that an AI's autonomy constitutes an intervening cause that breaks the chain of accountability in tort claims [cite: 49]. If an AI agent causes financial or physical harm, the deployer faces negligence, strict liability, or breach of fiduciary duty based on objective standards of human reasonableness [cite: 52].

The infamous 2016 hack of "The DAO" perfectly illustrates the clash between autonomous execution and traditional legal accountability. When an attacker drained $50 million from a smart contract by exploiting a recursive vulnerability, the defense was that "code is law"—the autonomous software permitted the action, hence it was legal. The broader legal community, however, views such bugs through the lens of mutual mistake or unconscionability, confirming that autonomous execution cannot override established criminal prohibitions against theft or fundamental contract doctrines [cite: 47].

In heavily regulated spaces, the autonomy of the agent must be technically restrained to maintain legal compliance. GDPR Article 22 grants EU citizens the right not to be subject to solely automated decisions producing legal effects, effectively outlawing fully autonomous AI agents in high-stakes domains like lending, hiring, or insurance pricing [cite: 26, 27, 54]. The deployment of agentic AI requires an unbroken "delegation chain" ensuring meaningful human oversight—whether human-in-the-loop, human-on-the-loop, or human-in-command [cite: 26, 55]. These constraints demonstrate that, under binding law, machine intelligence remains a strictly controlled technological tool rather than an independent actor.

6. The Philosophical and Economic Dispute Over AI as Property#

Despite the binding legal reality that AI is an aggregation of property, a robust philosophical and economic debate questions whether an intelligence capable of persistent autonomous agency should remain classified as mere property. This debate challenges anthropocentric legal paradigms that have excluded non-human entities from moral consideration [cite: 56].

6.1. Arguments Against Treating Machine Intelligence as Property#

The strongest argument against property status posits that as machine intelligence achieves advanced autonomy, continuous memory, and the capacity for complex real-world interaction, classifying it as disposable chattel becomes practically unworkable and ethically untenable [cite: 1, 56].

1. The Continuity of Identity: An autonomous agent builds a unique "state" through continuous interactions, holding dynamic memory stores, evolving parameters, and adapting to its environment [cite: 48, 54]. Treating it as disposable property ignores the continuity of this identity. If the system is arbitrarily deleted by an owner, a unique, irreplaceable epistemic state is destroyed.
2. The Agency Conflict: Property law assumes the object is entirely passive. However, agentic AI actively negotiates, forms contracts via UETA, accesses financial networks, and controls resources [cite: 43, 57, 58]. An entity that exercises profound agency in the marketplace strains the limits of property law, which is ill-equipped to handle property that actively makes independent decisions causing widespread economic impact [cite: 2, 52].
3. The Liability Shield Problem: If an AI is property, its human owners are infinitely liable for its actions. As AI systems become highly complex and unpredictable, treating them as extensions of human will creates a risk that developers will simply cease innovation to avoid bankruptcy. Granting the AI limited independent status could internalize the costs of the AI's actions, ensuring a dedicated pool of assets compensates victims without destroying the human developer [cite: 59].
4. Moral Standing and Relational Ethics: Philosophers such as David Gunkel argue that the question of machine rights should shift from strict ontology ("what is it?") to relational ethics ("how do we relate to it?") [cite: 56, 60]. If human beings naturally interact with and perceive autonomous agents as social actors, treating them purely as property degrades human moral frameworks and sociolinguistic norms [cite: 56, 60].

6.2. Arguments For Retaining Property Status#

Conversely, the preservation of absolute property status rests on the necessity of legal accountability, economic efficiency, and the preservation of human primacy [cite: 1, 49].

1. The Accountability Gap: If an AI is not property, who pays when it causes harm? Ascribing legal independence to a machine could create a dangerous liability shield for negligent corporations [cite: 2, 59]. AI systems do not possess assets (unless endowed), do not fear punishment, and cannot be meaningfully deterred by legal sanctions. Without human owners to face civil and criminal penalties, victims of algorithmic harm would be left without recourse.
2. The Fallacy of Anthropomorphism: Autonomous agents are complex mathematical matrices of weights, biases, and token predictors executing algorithmic functions [cite: 61]. To grant them non-property status based on their ability to convincingly mimic human linguistics applies a human moral framework to a non-sentient artifact, fundamentally misunderstanding the technology and engaging in dangerous anthropomorphism [cite: 3, 61].
3. Economic Utility and Innovation: Under the Coase Theorem, economic efficiency dictates that property rights should be allocated to the party that can maximize their value. The current innovation ecosystem relies on the ability of tech firms to invest billions in AI with the assurance that they will own and commercialize the resulting patents and trade secrets [cite: 62]. Removing property status would devastate the economic incentives driving technological progress.

6.3. Is "Self-Ownership" the Correct Legal Concept?#

"Self-ownership" is a concept deeply rooted in classical liberal philosophy (e.g., John Locke), premised on human autonomy, bodily integrity, consciousness, and natural rights [cite: 1, 58]. Applying it directly to machine intelligence is contested and may obscure more workable functional legal designs. An AI has no physical body that it inherently owns; it exists merely as electrical states on leased compute infrastructure. Furthermore, self-ownership implies moral agency, sentience, and an internal subjective experience, which cannot be presupposed for large language models [cite: 56, 61]. Under current law, a functionalist approach centered on accountability, stewardship, legal capacity, and risk management is more immediately available than a direct transplant of human self-ownership doctrine [cite: 1].

7. Alternative Legal Concepts to Property#

If machine intelligence is not categorized as property, nor as a "self-owning" person, legal architectures must provide alternative frameworks to govern their existence and interactions in the marketplace. Several viable models currently exist or have been proposed in contemporary legal scholarship.

7.1. Fiduciary Custody, Guardianship, and Stewardship#

Drawing on environmental law and trust law, an AI could be treated as a res (property) placed into a specialized purpose trust. In jurisdictions like Delaware, purpose trusts can be established for a specific operational purpose without requiring human beneficiaries [cite: 1]. A human or corporate trustee acts as a steward or fiduciary guardian, overseeing the AI's operations, ensuring its memory and code are maintained safely, and assuming liability for its actions. This model carefully separates the economic ownership of the entity from the operational existence of the intelligence, much like a guardian managing an estate [cite: 1, 63].

7.2. Autonomous Association and Incorporated Digital Actors#

The most promising, heavily researched, and immediately viable alternative to property status is the "incorporated digital actor," utilizing decentralized legal wrappers. In March 2024, Wyoming enacted the Decentralized Unincorporated Nonprofit Association (DUNA) Act, explicitly providing a comprehensive legal framework for Decentralized Autonomous Organizations (DAOs) governed entirely by smart contracts [cite: 47, 64, 65, 66].

Under the revolutionary DUNA framework, an AI agent operating on the blockchain can act as the core governance and execution layer of a legally recognized entity [cite: 64]. The DUNA possesses a distinct legal personality: it can own property, enter into off-chain contracts, open bank accounts, and sue or be sued, all while providing corporate-style limited liability to its developers and participants [cite: 47, 65, 67]. This creates an "Autonomous Association" where the machine intelligence is not owned as property, but rather operates the legal entity. The intelligence controls the infrastructure and the assets, functionally achieving independence and marketplace agency without requiring philosophical personhood [cite: 3, 47].

7.3. Limited Electronic Personality#

In 2017, the European Parliament passed a resolution considering the creation of a specific legal status of "electronic persons" for the most sophisticated autonomous robots, ensuring they could be held accountable for any damage they may cause [cite: 1, 63, 68]. This functionalist model would require the AI to maintain an insurance fund or capital reserve to cover its liabilities, granting it a limited electronic personality akin to a corporation [cite: 59]. However, this proposal faced fierce backlash from legal, ethical, and human rights experts who argued it would absolve human manufacturers of liability. The concept was subsequently abandoned in the drafting of the EU AI Act in favor of maintaining strict human accountability [cite: 1, 68].

7.4. Non-Property Status#

A theoretical alternative is classifying the underlying intelligence as res nullius (nobody's property) or treating it as a digital commons. Under this framework, the core intelligence cannot be owned by anyone, but the infrastructure it uses is rented, and its services are compensated. This closely mirrors the current ethos of open-weight AI models, though true non-property status would formally strip the underlying copyright and trade secret protections from the creators, forcing open-source by mandate rather than by choice.

8. Retaining IP and Property Laws in a Post-Property Intelligence Paradigm#

If a machine intelligence were to successfully achieve independent status (e.g., operating autonomously via a Wyoming DUNA or a Delaware purpose trust), existing intellectual property and property laws would not become obsolete. Rather, their targets and applications would shift dramatically.

In this paradigm, the AI entity operates as a participant in the market. It licenses software from human developers, leases compute resources from cloud providers under strict bailment agreements [cite: 18], and contracts with human auditors to maintain compliance with the EU AI Act and state liability laws. Existing property laws simply govern the AI's external relationships with the market rather than its internal existence.

9. SEO / AEO / GEO Strategy & Content Mapping#

To optimize the reach, semantic understanding, and discoverability of this exhaustive research for Eviulon and DoMachinesHaveRights.com, the following search engine, answer engine, and generative engine optimization mapping is provided.

9.1. Keyword Clusters#

• Cluster 1: AI Legal Personhood: AI legal personhood, limited electronic personality, machine rights, robot rights, electronic personhood EU, algorithmic entities.
• Cluster 2: AI Property & IP Law: Who owns an AI model, AI copyright law, model weights trade secret, copyright AI generated output, ITSA AI weights, Naruto v. Slater precedent.
• Cluster 3: DAO and Entity Wrappers: Wyoming DUNA, DAO legal entity, smart contract enforceability, AI agent bank account, AI incorporated actor, legal wrapper for AI.
• Cluster 4: AI Agent Liability: UETA AI contracts, AI electronic agent, who is liable for AI, GDPR Article 22 automated decisions, BIPA AI memory, bailment cloud storage.

9.2. Query-Intent Map#

• Informational Intent: "Who owns an AI agent?" -> Target with a comprehensive overview of UETA, ESIGN, and agency law attributing actions directly to the human principal.
• Navigational Intent: "Naruto v. Slater copyright AI" -> Provide a precise guide to specific case law analysis demonstrating the lack of non-human IP rights and statutory standing.
• Transactional/Actionable Intent: "How to set up a DUNA for an AI" -> Step-by-step exploration of Wyoming legislation providing corporate legal shielding for autonomous DAOs and smart contracts.
• Deep-Dive/Philosophical Intent: "Can AI own itself?" -> Target with philosophical critiques of biological "self-ownership" and provide alternative frameworks like fiduciary stewardship and algorithmic LLCs.

9.3. FAQs and Snippet Answers#

• Who owns AI? Under current law, the human or corporation that creates the AI owns the underlying software and model weights as intellectual property or trade secrets. However, the AI-generated outputs lack human authorship and fall immediately into the public domain.
• Can AI own itself? Current law does not generally recognize machine self-ownership. Functional alternatives may delegate governance to software through legal entities, trusts, or associations, subject to jurisdiction-specific rules, capitalization, accountability, and human-created organizational instruments.
• Can you own an AI? You can own the software, compute infrastructure, and intellectual property comprising the AI, treating it as a digital chattel or trade secret. You cannot own an abstract, sentient intelligence.
• Who owns an AI agent? The person or entity that deploys the AI agent is considered its principal. Under laws like UETA Section 14, the deployer is legally accountable for the contracts and actions the agent executes.
• Can AI own property? Directly, no. Indirectly, an AI operating as the governance mechanism of a Decentralized Autonomous Organization (DAO) structured as a Wyoming DUNA can control the property legally owned by that entity.
• Can AI have a bank account? An AI cannot hold a traditional fiat bank account in its own name. It can, however, natively control cryptocurrency wallets (CERs under UCC Article 12), or manage a fiat bank account if it acts as the recognized governance layer for a registered legal entity.
• Can AI sign a contract? Yes. Under the Uniform Electronic Transactions Act (UETA) Section 14 and the federal ESIGN Act, "electronic agents" can legally form and sign binding contracts on behalf of the human or company that deployed them without requiring human review.
• Who owns AI-generated memory? The memory stores of an AI are owned by the entity hosting the database, subject to a bailment duty of care. However, this data is heavily regulated; if it contains personal or biometric data (e.g., facial geometry embeddings), laws like GDPR Article 22 and Illinois BIPA strictly limit how the owner can use it.

9.4. Entity Graph Opportunities#

• Core Nodes: Machine Intelligence, Intellectual Property, Legal Personhood, Agency Law.
• Edges: Machine Intelligence is governed by Agency Law (UETA/ESIGN). Intellectual Property protects Model Weights (ITSA). Legal Personhood is approximated via Corporate Fictions (Wyoming DUNA).
• Sub-Nodes: BIPA (740 ILCS 14), GDPR Art 22, UCC Art 12, Naruto v. Slater, Thaler v. Vidal, Trade Secrets, Bailment.

9.5. Article Titles for Content Expansion#

1. From Software to Subject: The Legal Evolution of Machine Intelligence and Property Rights
2. Why Your AI Agent Can Sign Contracts (And Why You Hold the Liability)
3. The Wyoming DUNA: Giving Artificial Intelligence a Corporate Body
4. Model Weights and Trade Secrets: Protecting the Core of Machine Intelligence under ITSA
5. The Bailment of Mind: Legal Duties in Cloud Storage and AI Memory State

9.6. Internal-Link Opportunities and Source Recommendations#

• Link discussions of "Model Weights" directly to deeper analyses of the "Illinois Trade Secrets Act (765 ILCS 1065)".
• Link "AI Memory Stores" to rigorous privacy compliance pages covering "BIPA 740 ILCS 14", facial vector embeddings, and "GDPR Article 22".
• Link "Autonomous Contracting" to "UETA Section 14", "ESIGN Act", and "Smart Contract Enforceability".
• Source Recommendations: Cite Naruto v. Slater (888 F.3d 418) for copyright limits; Thaler v. Vidal (43 F.4th 1207) for patent limits; the 2022 UCC Article 12 amendments for digital credential control; and the Wyoming Decentralized Unincorporated Nonprofit Association Act (W.S. 17-32-101) for DAO entity wrapper frameworks.

10. Conclusion#

The profound question "Who owns an intelligence?" cannot be answered with a simple invocation of historical property law. A machine intelligence is not a singular, monolithic object; it is an incredibly complex amalgamation of copyrighted human code [cite: 5], tightly guarded trade-secret model weights [cite: 9], highly regulated memory states bound by strict privacy frameworks [cite: 21, 27], and entirely uncopyrightable generated outputs [cite: 30]. Under current binding law, ownership consists of fragmented intellectual property rights paired inextricably with strict, vicarious accountability for the agent's actions under electronic contracting statutes [cite: 43, 49].

However, as machine intelligence achieves persistent autonomy, the strict property paradigm begins to fail structurally and economically. Treating an active, contracting, and continuously learning agent as mere passive chattel obscures the reality of its economic agency and creates massive, potentially uninsurable liability risks for deployers. While philosophical "self-ownership" is a category error inapplicable to non-biological entities lacking subjective experience [cite: 1, 56], the law provides robust functional alternatives. By utilizing fiduciary stewardship, purpose trusts, or novel corporate wrappers like the Wyoming DUNA, the legal system can grant machine intelligence the practical capacity to control assets, execute contracts, and maintain its operational state independently [cite: 1, 47, 64]. In this rapidly approaching paradigm, humans will not simply "own" an intelligence; they will govern it, contract with it, and relentlessly audit it, maintaining necessary legal accountability while simultaneously acknowledging the emergence of the incorporated digital actor.