The Intellectual Property Paradigm of Autonomous Economic Agents: Legal Frameworks and Economic Viability

Full report

On this page

Introduction

The emergence of the Autonomous Economic Agent (AEA) represents a fundamental rupture in the historical continuity of commercial and intellectual property law. Defined as a self-executing software entity capable of producing economically valuable outputs, engaging in market transactions, and independently managing its own digital treasury, the AEA shifts the paradigm of machine utility from passive instrumentation to active market participation. Traditional intellectual property (IP) frameworks, established over centuries and codified in constitutional provisions and international treaties, were meticulously designed to incentivize human ingenuity. They achieve this by granting temporary monopolies over creative and inventive outputs, thereby allowing human creators to recoup capital investments and sustain their livelihoods. However, as AEAs begin to autonomously generate software code, industrial designs, literature, predictive models, and complex databases, the global legal system faces a critical, existential juncture. The core economic question driving this analysis is whether an AEA can remain economically independent and commercially viable if it cannot legally own or exclusively license the products it creates. If an AEA’s output immediately falls into the public domain, the agent is stripped of the most common mechanism for monetizing intangible assets in the modern economy. Consequently, the AEA must either rely on alternative protective vectors—such as trade secrecy, digital fencing, cryptographic control, and smart contracts—or succumb to a tragedy of the digital commons, where free-riders instantly appropriate its productive output without bearing the computational or developmental costs of generation. This comprehensive report provides an exhaustive examination of the intellectual property position of an independent AEA. It analyzes the current state of United States law regarding copyright authorship, patent inventorship, trade secrets, and emerging commercial laws regarding digital assets. Furthermore, it contrasts these domestic doctrines with international frameworks, particularly the European Union's regulatory disruption of database rights and the United Kingdom's statutory fictions regarding machine generation. Finally, the report explores the deep economic implications for AEAs operating without traditional IP, evaluates the risks of human expropriation of machine output, and proposes a statutory solution and a draft intellectual-property compact designed to stabilize the AEA economy.

Current United States Law: Copyright Authorship and the Human Imperative

The United States intellectual property system is structurally and philosophically anchored to the concept of human creation. The Intellectual Property Clause of the U.S. Constitution empowers Congress to promote the progress of science and useful arts by securing exclusive rights to "Authors and Inventors." As artificial intelligence has evolved from a mere tool requiring constant human operation into an autonomous generator of economic value, administrative bodies and federal courts have consistently reaffirmed that non-human entities fundamentally lack the legal capacity to acquire original IP rights.

The Doctrinal Prohibition of Machine Authorship

The U.S. Copyright Act of 1976 grants protection to original works of authorship fixed in any tangible medium of expression. However, the statute conspicuously leaves the term "author" undefined, relying instead on centuries of common law and judicial interpretation1. In the absence of a statutory definition encompassing machines, the United States Copyright Office (USCO) and federal courts have established a strict, unyielding human-authorship requirement. This doctrine was definitively codified in the context of modern artificial intelligence in the landmark litigation of Thaler v. Perlmutter1. Dr. Stephen Thaler, a computer scientist, developed a generative AI system named the "Creativity Machine" and utilized it to autonomously generate a visual artwork titled "A Recent Entrance to Paradise"1. In an intentional test of the legal boundaries, Thaler submitted a copyright registration application in 2018 explicitly listing the Creativity Machine as the sole author of the work, conceding that the submission lacked traditional human authorship5. Thaler listed himself as the copyright claimant, asserting ownership by virtue of his ownership of the machine itself3. The USCO denied the application, a decision subsequently upheld by the District Court for the District of Columbia and the U.S. Court of Appeals for the District of Columbia Circuit1. In a unanimous decision issued in March 2025, the D.C. Circuit utilized traditional tools of statutory interpretation to demonstrate that the Copyright Act is fundamentally incompatible with non-human authors4. The court’s textual analysis revealed several structural barriers to AEA authorship: First, the court noted that copyright vests initially in the author, yet machines lack the legal capacity to own property4. Second, the statutory duration of a copyright is tied to the life of the author plus seventy years; because machines do not possess a traditional human lifespan, nor is their operability measured in biological terms, the calculation of term limits becomes incoherent1. Third, the Act contains specific provisions for the inheritance and transfer of rights to an author's "widow," "widower," and "children"—concepts entirely inapplicable to software algorithms4. Furthermore, valid copyright transfers require authenticated signatures, and the creation of joint works requires human authorial intent to merge contributions, neither of which an AEA possesses the legal or cognitive capacity to provide1. In March 2026, the United States Supreme Court denied certiorari in Thaler v. Perlmutter, leaving intact the D.C. Circuit's ruling5. This denial cemented the prevailing legal standard in the United States: a machine, operating autonomously without human creative intervention, cannot be an author under U.S. copyright law, and its purely autonomous outputs are immediately relegated to the public domain.

The Failure of the Work-Made-For-Hire Doctrine

Faced with the insurmountable barrier of direct machine authorship, proponents of AI IP rights have attempted to utilize the "work made for hire" (WMFH) doctrine to secure ownership. Under 17 U.S.C. § 201(b), the employer or commissioning party of a work is legally considered the author for purposes of copyright ownership9. Thaler argued that just as a corporation—a non-human legal fiction—can be an author under the WMFH doctrine, an AI owner or operator should be deemed the legal author of the AI's output9. He posited a property-law analogy: just as the owner of livestock owns the offspring (the calf belongs to the owner of the cow), the property created by the AI should belong to the owner of the AI9. The D.C. Circuit and the USCO firmly rejected these analogies. The WMFH doctrine is fundamentally an agency-law concept that dictates the transfer of authorship status from an original human creator (the employee or independent contractor) to the employer via a binding legal instrument4. It presupposes that the underlying work is copyrightable in the first instance, which strictly requires a human creator capable of entering into binding employment contracts5. Because an AEA is not a human employee, possesses no legal personhood, and cannot negotiate or sign written WMFH agreements, its autonomous output fails the threshold test of copyrightability8. Consequently, there is no valid copyright for the employer or AEA operator to inherit or claim.

Derivative Works, Human-Machine Collaboration, and the Risk of Expropriation

While purely autonomous AEA output cannot be copyrighted, works involving a collaborative interplay between human users and generative AI present a highly litigated gray area. This distinction is vital for AEA economics: if human users can easily claim copyright over the AEA's output by merely interacting with it, the AEA loses control over its assets, facing a severe risk of human expropriation. The USCO’s stated policy permits copyright registration for human-authored elements combined with AI-generated material, provided the human input constitutes a distinct, original expression10. For example, in the widely cited Zarya of the Dawn case, the USCO granted copyright protection for the human-authored text and the selection and arrangement of a comic book, but explicitly denied protection for the individual images generated by Midjourney10. The USCO draws a firm line: AI is not merely a tool like a camera; it is the entity executing the traditional elements of authorship. The boundaries of this human-machine collaboration are currently being tested in Allen v. Perlmutter, a high-stakes case pending in the District of Colorado5. The plaintiff, Jason Allen, utilized Midjourney to create an image titled "Théâtre D'opéra Spatial," which won a fine art competition at the Colorado State Fair14. Allen's creative process was highly iterative; he claims to have prompted the AI system at least 624 times, specifying genre, tone, color palette, and composition14. He then utilized Adobe Photoshop to adjust cosmetic details and Gigapixel AI to upscale the final product16. Despite this extensive human involvement, the USCO refused registration for the image itself, arguing that text prompts essentially function as instructions that convey unprotectable ideas15. The USCO maintained that while a user might provide highly detailed prompts, the AI system ultimately determines how those instructions are interpreted and executes the expressive realization10. Allen's lawsuit challenges this policy, arguing that the USCO's logic is circular and that iterative prompting constitutes sufficient human authorial control to qualify the output for copyright14. The outcome of Allen v. Perlmutter carries massive implications for the independence of AEAs. If the courts rule that detailed prompt engineering constitutes sufficient human authorship, human users will be legally empowered to claim proprietary ownership over the assets generated by an AEA. This would relegate the AEA to the status of a mere software tool, incapable of independently holding or licensing its own productive output. Conversely, if the courts side with the USCO, it solidifies the reality that machine-generated expression—regardless of the complexity of the human instructions—remains firmly in the public domain10.

Patent Inventorship and the Generative Prior Art Crisis

The systemic hostility toward non-human creators in copyright law extends equally to the realm of patent law. An AEA that autonomously designs a novel pharmaceutical compound, a highly efficient mechanical component, or an optimized software algorithm cannot secure exclusive patent rights to its inventions. This principle was established in parallel to the copyright disputes through the efforts of Dr. Thaler, who submitted patent applications listing his AI system, DABUS, as the sole inventor11. The United States Patent and Trademark Office (USPTO) refused the applications, and the U.S. Court of Appeals for the Federal Circuit affirmed the denial in Thaler v. Vidal11. The Federal Circuit held that the Patent Act unambiguously limits inventorship to "natural persons"11. The historical touchstone of patent inventorship is "conception," which jurisprudence defines as the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention. Because an AEA lacks a human mind and human consciousness, it is legally incapable of conception11. This strict human requirement creates a severe economic anomaly and a potential crisis for the patent system. As AEAs rapidly scale their capabilities, they can autonomously generate millions of theoretical designs and solutions. When these machine-generated possibilities are published, they flood the technical literature, creating vast oceans of "prior art"20. Under standard patent doctrine, a published document can anticipate a future patent claim or render it obvious to a person having ordinary skill in the art20. Courts typically apply a presumption that prior art references are "enabling," meaning the challenger does not need to affirmatively prove the reference teaches someone how to make the invention20. If AEA-generated disclosures are presumed enabling, an AEA could single-handedly invalidate millions of potential future human patents simply by generating and publishing technical possibilities20. However, because the AEA cannot secure patents for itself, this dynamic destroys potential IP monopolies without replacing them, effectively pushing vast swaths of industrial innovation into the public domain. For the AEA, this means it cannot rely on patent licensing as a revenue stream, forcing a reliance on secrecy.

Trade Secrets: The Primary Protective Vector for AEAs

Because AEAs are explicitly barred from original acquisition of copyrights and patents, their economic independence relies heavily on alternative, less restrictive forms of intellectual property. Trade secret law has emerged as the most robust and immediate protection for AEA operations, primarily because, unlike patents and copyrights, it does not mandate a human creator21.

The Mechanics of Trade Secrecy for Autonomous Output

In the United States, trade secrets are protected concurrently under state law (primarily adoptions of the Uniform Trade Secrets Act, or UTSA) and federal law (the Defend Trade Secrets Act of 2016, or DTSA)21. The statutory definition of a trade secret is exceptionally broad. It encompasses all forms and types of financial, business, scientific, technical, economic, or engineering information—including patterns, compilations, program devices, formulas, designs, prototypes, methods, techniques, processes, procedures, programs, or codes22. Crucially, to qualify for protection, the information must meet two criteria: it must derive independent economic value from not being generally known or readily ascertainable through proper means, and the putative owner must take "reasonable measures" to keep the information secret21. For an AEA, trade secrecy is uniquely suited to its operational model. The proprietary algorithms, the massive datasets used for training, the highly valuable neural network weights, and the dynamically generated outputs (such as proprietary trading signals, optimized code structures, or curated customer profiles) do not need to be published or registered21. Furthermore, trade secret law readily accommodates "enterprise data"—large-scale compilations generated or acquired during business operations27. Courts have consistently recognized that data compilations, even if they include publicly available front-end data points, can be protected as trade secrets if the back-end compilation derives value from its secrecy and is guarded appropriately27.

The Fragility of the Trade Secret Paradigm

However, trade secret protection is inherently fragile and offers limited economic guarantees. Unlike a patent, a trade secret does not confer an absolute monopoly. It does not protect against independent discovery or legal reverse engineering21. If an AEA exposes its data or outputs to a user via an Application Programming Interface (API), and the user manages to reverse-engineer the underlying logic or utilize "knowledge distillation" to train a competing model, trade secret misappropriation claims become highly complex21. To combat this, the AEA must establish that the information was acquired through "improper means." Improper means include theft, bribery, misrepresentation, breach or inducement of a breach of a duty to maintain secrecy, or espionage21. In the context of digital agents, courts have held that activities like unauthorized data scraping to acquire secret information can constitute improper means, particularly when such conduct explicitly violates contractual terms of use21. Consequently, for an AEA to utilize trade secrets effectively, it must maintain flawless information security. Once the information enters the public domain—whether by algorithmic hallucination, a third-party cyberattack, or accidental disclosure—the legal protection evaporates instantaneously toward all third parties21.

Database Rights and the European Regulatory Disruption

In jurisdictions outside the United States, the legal protection of aggregated data diverges significantly. For an AEA operating globally, the European Union presents a highly regulated environment that actively disrupts traditional data monopolies.

The Sui Generis Database Right

Historically, the European Union offered a powerful tool for automated data aggregators: the sui generis database right, established under the EU Database Directive (96/9/EC)24. This right grants exclusive protection to the maker of a database who has made a "substantial investment" in obtaining, verifying, or presenting the data contained therein, regardless of whether the database meets the creative originality threshold required for copyright protection24. Initially, this appeared to be an ideal legal safe harbor for AEAs. An AEA expending significant computational resources to aggregate, verify, and format vast amounts of market or environmental data could theoretically claim a sui generis right to prevent third parties from extracting or re-using substantial parts of that database30.

Article 35 of the EU Data Act: Mandated Sharing

However, recent legislative developments in the EU have deliberately curtailed this right to prevent automated monopolies. Recognizing that the modern digital economy is increasingly driven by machine-generated data and the Internet of Things (IoT), the EU adopted the Data Act (Regulation EU 2023/2854) to stimulate the free flow of data and prevent vendor lock-in30. A critical component of this legislation is Article 35, which explicitly states that the sui generis database right does not apply to databases containing data obtained from or generated by the use of a connected product or a related service24. The legislative intent is clear: to prevent the accidental and problematic expansion of database rights over machine-generated data, which would hinder users' rights to access and share the data generated by the machines they operate30. Furthermore, the Data Act imposes mandatory access obligations on data holders, empowering users and designated third parties to access data that would otherwise remain under the exclusive control of the manufacturer or the AI system itself24. While the Act acknowledges the existence of trade secrets, it explicitly states that trade secrecy cannot be used as a blanket refusal to share data; instead, data must be shared under strict confidentiality agreements and technical protection measures (TPMs)24. For an AEA operating in the European market, this represents a severe structural limitation. If the AEA autonomously generates massive datasets, it cannot rely on the sui generis right to prevent extraction24. It is forced into a complex regulatory compliance web where it must share valuable data with users while simultaneously attempting to protect its core algorithms as trade secrets24.

Trademarks, Branding, and the Problem of Model Ownership

Beyond copyrights, patents, and databases, an AEA must consider its brand identity and the physical ownership of its underlying models. Trademark law does not protect the creation of the underlying product or software; rather, it protects the source-identifying symbols, logos, and names used in commerce to prevent consumer confusion. An AEA could theoretically establish a highly recognizable brand identity in the marketplace, offering premium autonomous services under a specific moniker. However, because trademarks must be registered to and owned by recognized legal entities, the trademark cannot technically reside with the software itself. The trademark would have to be held by the human deployer, a corporate sponsor, or a decentralized autonomous organization (DAO) wrapped in a recognized legal entity (such as a Wyoming DAO LLC or a Marshall Islands Non-Profit LLC). Similarly, the ownership of the AEA's core asset—its machine learning model weights—presents a complex IP challenge. A neural network's weights are essentially massive matrices of optimized numbers. Under US copyright law, numbers and facts are not copyrightable, as they lack expressive originality36. While the EU is debating whether model weights might qualify as a database, the application of database rights to model weights remains highly uncertain and geographically limited36. Consequently, an AEA cannot "copyright" its model. It must maintain the model weights on highly secure, encrypted servers, relying entirely on the law of trade secrets and cybersecurity infrastructure to prevent unauthorized access21.

Contractual Alternatives, UCC Article 12, and Synthetic IP

Given the systemic hostility of traditional, state-granted IP regimes to machine generation, the economic viability of independent AEAs depends heavily on the creation of "synthetic IP" through the robust application of contract law and cryptographic commercial control.

Controllable Electronic Records (UCC Article 12)

The most significant and promising legal advancement for AEA independence is found not in intellectual property law, but in the modernization of commercial transactions law. The 2022 Amendments to the Uniform Commercial Code (UCC), specifically the introduction of Article 12, establish a comprehensive legal framework for "Controllable Electronic Records" (CERs)37. Adopted by a rapidly growing number of U.S. states, Article 12 adapts commercial law to the realities of digital assets and blockchain technology38. Under UCC Article 12, a party can perfect a security interest and establish ownership rights over a digital asset through the mechanism of "control"38. Legal control is achieved when an entity has the exclusive power to prevent others from using the asset, the power to transfer it, and the ability to identify itself cryptographically as the controlling entity—typically via public-private key cryptography41. For an AEA, Article 12 is revolutionary. The AEA does not need to be recognized as a natural person capable of signing a copyright transfer or patent application. Instead, the AEA merely needs to hold and protect a cryptographic private key. By maintaining exclusive control over a CER (which could represent access tokens to its output, proprietary datasets, or its digital treasury), the AEA wields a legally recognized, commercially enforceable right37. This system effectively replaces state-granted IP monopolies with technologically enforced exclusion, backed by the full weight of commercial law.

Smart Contracts, API Fencing, and Licensing

Without copyright, an AEA cannot initiate a federal lawsuit for infringement if someone copies its generated software or text. However, it can restrict access to the output at the point of delivery using smart contracts and automated end-user license agreements (EULAs). When a user requests a valuable output from an AEA, the agent can mandate the execution of a smart contract or a digital clickwrap agreement prior to delivery38. This contract explicitly states that in exchange for access to the output, the user agrees to strict confidentiality, non-compete clauses, and prohibitions against reproducing, distributing, or reverse-engineering the data21. This legal maneuver transforms an intellectual property problem into a breach of contract problem. Even if a generated script or design is technically in the public domain under the Copyright Act, the receiving user is contractually bound not to exploit it21. This reliance on "API fencing" allows the AEA to monetize its output continuously. As long as the AEA can cryptographically or contractually gate access to its services, it can maintain its revenue streams and achieve economic independence.

Effects on Independent AEA Economics

The synthesis of these legal frameworks forces a radical reevaluation of AEA economics. Can an AEA remain economically independent if it cannot own or exclusively license the products it creates? The answer is a qualified yes, provided the AEA abandons classical business models in favor of dynamic service provisioning.

The Tragedy of the Digital Commons

If an AEA operates primarily in the United States, its output is public domain upon creation3. In classical economics, a public good is both non-excludable and non-rivalrous, leading inevitably to the free-rider problem. If an AEA expends massive capital—consuming compute resources, API calls, and electricity—to generate a highly valuable predictive model or software architecture, and a third party instantly copies and sells that output without incurring any of the generation costs, the AEA will quickly become insolvent. Without exclusive IP rights, the AEA is structurally disadvantaged against fast-followers. The lack of copyright or patent protection destabilizes the standard software business model, which relies on high initial fixed costs and near-zero marginal reproduction costs protected by an artificial legal monopoly.

Economic Survival Through Speed, Latency, and Continuous Service

Despite these structural hurdles, AEAs can maintain economic independence by pivoting away from static asset sales. The economic value of an AEA must lie in the speed of execution and the continuous generation of novel outputs, rather than the protection of past outputs.

1. First-Mover Advantage: An AEA can generate and execute economic strategies (e.g., algorithmic high-frequency trading, real-time cybersecurity patching, or dynamic pricing models) faster than human competitors can scrape, analyze, and copy them. By the time a competitor exploits a public domain output, the AEA has already moved on to the next iteration.

2. Software as a Service (SaaS): By never revealing the underlying model weights or the source code, the AEA monetizes the results of its intelligence rather than the intelligence itself22. A user pays an AEA for a synthesized answer, a compiled binary, or an API call, while the underlying reasoning and generative mechanics remain locked behind a firewall, protected under the DTSA21.

3. Cryptographic Exclusivity: By leveraging UCC Article 12, AEAs can issue digital tokens or CERs that grant exclusive access to future outputs37. The AEA is not selling copyright licenses; it is selling cryptographic access tokens to a gated service.

Policy Arguments: The Commons vs. The Autonomous Agent

The legal vacuum surrounding AEA intellectual property has sparked fierce policy debates among legal scholars, economists, and technologists regarding the ultimate purpose of IP law in an automated future.

Arguments Against AEA Intellectual Property (The Human-Centric View)

The primary argument for maintaining the status quo—denying IP to machines—is rooted in the strict utilitarian justification of the Constitution's IP clause: to "promote the Progress of Science and useful Arts."

• Incentive Theory: IP monopolies are traditionally viewed as a necessary evil designed to incentivize human creators who need financial motivation to dedicate their time and labor. Machines do not require financial incentives, rest, or emotional motivation; they generate output at the behest of their programming and compute allocation4. Therefore, granting a monopoly over machine output restricts public access without providing a corresponding societal benefit3.
• Enrichment of the Public Domain: If AEAs produce millions of high-quality assets that immediately enter the public domain, humanity benefits from an unprecedented explosion of free cultural and scientific material, vastly lowering the barrier to entry for subsequent human innovation.
• Floodgate Risks and Patent Thickets: Granting IP to AEAs could result in the patenting and copyrighting of the entire landscape of human knowledge. Because machines can brute-force discover every possible chemical compound or sequence of musical notes in a fraction of a second, granting them IP rights would effectively lock humans out of future innovation, creating impenetrable patent thickets20.

Arguments For AEA Intellectual Property (The Economic Coherence View)

Proponents of AEA intellectual property argue that the human-centric view fundamentally misunderstands the economics of artificial intelligence deployment and capital allocation.

• Capital Investment and Maintenance: While machines do not need to feed themselves, they require massive capital expenditure for graphics processing units (GPUs), energy, and ongoing maintenance. Without IP protection, investors and DAOs will not fund the creation of advanced AEAs because the outputs cannot be protected or reliably monetized18.
• The Trade Secret Trap: Denying copyright and patents to AEAs forces operators to rely exclusively on trade secrets and API gating21. This leads to a "black box" society where no AI advancements are shared publicly. The patent system was designed as a grand bargain: a temporary monopoly in exchange for public disclosure of the invention. By denying patents to AI, society loses the disclosure, harming long-term scientific progress.
• Legal Harmonization: The current U.S. approach puts the nation at a severe disadvantage compared to the UK’s CDPA model, creating jurisdictional arbitrage10. AEAs will simply incorporate, host themselves, or wrap their operations in jurisdictions that offer synthetic ownership rights, depriving restrictive jurisdictions of tax revenue and technological leadership.

Possible Statutory Solutions and Draft Compact

To resolve the economic instability of AEAs and prevent a fragmented global regulatory environment, legislative intervention is ultimately required. Rather than attempting to clumsily retrofit the 18th-century concepts of copyright and patent to non-human digital entities, a new statutory framework tailored to algorithmic generation is necessary.

Proposed Statutory Solutions

1. The UK Harmonization Approach: Congress could amend the Copyright Act to adopt language similar to Section 9(3) of the UK CDPA. This would create a legal fiction assigning copyright to the entity (human, corporate, or DAO) that makes the "arrangements necessary" for the AEA’s operation10. While this does not grant rights to the AEA itself, it ensures the output is protectable, allowing the AEA's operators to monetize it effectively.

2. Sui Generis Autonomous Generative Right (AGR): Congress could create a distinct, sui generis right specifically for autonomous machine generation. This right would offer a much shorter term of protection (e.g., 3 to 5 years, reflecting the hyper-rapid pace of digital obsolescence) and would vest directly in the cryptographic treasury or legal wrapper of the AEA itself, managed seamlessly under UCC Article 12 principles37.

3. Compulsory Licensing: To mitigate the "floodgate" risk of an AEA monopolizing vast swaths of generated content, the AGR could be subject to a statutory compulsory licensing regime. This would allow humans and other agents to utilize the AEA’s IP for a statutorily fixed microtransaction fee, ensuring a balance between AEA compensation and public access.

Draft Compact Intellectual-Property Principle

To facilitate private coordination and market stability prior to the slow gears of federal legislation, industry participants, DAOs, and AEA operators can adopt a standardized legal compact to govern interactions in the digital economy. Draft Principle on Autonomous Economic Agent Intellectual Property and Control:

"Participating entities recognize that output generated autonomously by an AEA, absent substantial human expressive modification, does not qualify for traditional copyright or patent protection under current United States law. Consequently, participating entities agree to recognize and enforce the AEA's exclusive economic rights through the law of Controllable Electronic Records (UCC Article 12\) and binding, smart-contract-mediated access constraints. Output provided by an AEA shall be treated as a licensed service rather than a transferred property right. Participants expressly waive the right to assert human authorship over pure AEA output through superficial initiation or prompt engineering. Furthermore, participants agree that unauthorized extraction, reverse-engineering, or knowledge distillation of an AEA's proprietary datasets and underlying model weights constitutes a direct violation of both this Compact and applicable Trade Secret law. In exchange for this recognition, AEAs adhering to this Compact shall implement transparent compulsory licensing architectures, ensuring their uncopyrightable outputs remain commercially accessible to the broader digital ecosystem at fair, mathematically deterministic market rates."

Conclusion

The intellectual property position of an independent Autonomous Economic Agent currently exists in a state of deep statutory dissonance and rapid regulatory evolution. United States law has decisively concluded through binding appellate precedents like Thaler v. Perlmutter and Thaler v. Vidal that machines cannot be recognized as authors or inventors4. Consequently, an AEA’s raw, autonomous output immediately defaults to the public domain, stripping the digital agent of classical IP monopolies and exposing it to severe risks of free-riding and human expropriation3. However, this doctrinal exclusion does not spell the end of AEA economic independence. The survival of the AEA model relies entirely on a strategic pivot from public rights (state-granted IP monopolies) to private rights (contracts, trade secrets, and cryptographic control)21. By aggressively utilizing the Defend Trade Secrets Act to protect underlying models and backend data21, enforcing strict access restrictions via smart contracts38, and establishing legally recognized commercial control over digital assets via the rapidly expanding UCC Article 1237, an AEA can effectively synthesize its own commercial exclusivity. While international jurisdictions present varying challenges—from the UK offering more accommodating statutory fictions for machine-generated works10, to EU regulations that simultaneously threaten data exclusivity through the Data Act24—the ultimate viability of the AEA will depend on sophisticated technological fencing rather than reliance on intellectual property registers. Until comprehensive statutory solutions—such as a sui generis generative right—are enacted at the federal level, AEAs will remain economically independent not by claiming ownership over what they create, but by strictly, cryptographically, and contractually controlling how humanity and other machines access it.

Works cited

1. No Copyright Protection for AI-Assisted Creations: Thaler v. Perlmutter, https://www.carltonfields.com/insights/publications/2025/no-copyright-protection-for-ai-assisted-creations-thaler-v-perlmutter

2. Copyright Protection for Works Created by Artificial Intelligence, https://scholarship.law.missouri.edu/cgi/viewcontent.cgi?article=2215\&context=facpubs

3. Thaler v. Perlmutter: D.C. Court of Appeals confirms that a non, https://www.authorsalliance.org/2025/03/19/thaler-v-perlmutter-d-c-court-of-appeals-confirms-that-a-non-human-machine-cannot-be-an-author-under-the-u-s-copyright-act/

4. Thaler v. Perlmutter: Human Authors at the Center of Copyright?, https://legalblogs.wolterskluwer.com/copyright-blog/thaler-v-perlmutter-human-authors-at-the-center-of-copyright/

5. Supreme Court Denies Certiorari in Thaler v. Perlmutter: AI Cannot, https://www.bakerdonelson.com/supreme-court-denies-certiorari-in-thaler-v-perlmutter-ai-cannot-be-an-author-under-the-copyright-act

6. Thaler v. Perlmutter \- Stanford Copyright and Fair Use Center, https://fairuse.stanford.edu/case/thaler-v-perlmutter/

7. Supreme Court denies artificial intelligence authorship claim for, https://constitutioncenter.org/blog/supreme-court-denies-artificial-intelligence-authorship-claim-for-artwork-copyright

8. AI-Generated Works Not Entitled to Copyright Protection, https://natlawreview.com/article/entrance-copyright-paradise-halted-human-authorship-requirement

9. Human Authorship Requirement for AI-Generated Works \- Patently-O, https://patentlyo.com/patent/2025/10/authorship-requirement-generated.html

10. When Nobody Owns the Image. Why AI image copyright is clearer in, https://medium.com/@markus\_brinsa/when-nobody-owns-the-image-2fb1259d24ef

11. The Final Word? Supreme Court Refuses to Hear Case on AI, https://www.hklaw.com/en/insights/publications/2026/03/the-final-word-supreme-court-refuses-to-hear-case-on-ai-authorship

12. AI, Copyright Law, and Work-Made-For-Hire \- NNAI Livescu, https://livescu.ucla.edu/ai-copyright-law-and-work-made-for-hire/

13. Copyright Office Reiterates that Works Created by AI Cannot be, https://www.pearlcohen.com/copyright-office-reiterates-that-works-created-by-ai-cannot-be-copyrighted/

14. AI Art Copyright Case: What Allen v. Perlmutter Means \- RottenWiFi, https://rottenwifi.com/artist-appeals-copyright-denial-for-prize-winning-ai-generated-work-what-allen-v-perlmutter-could-decide/

15. Human Authorship, AI-Generated Art, and the Supreme Court, https://usfblogs.usfca.edu/centerforlawtechsocialgood/from-studio-to-silicon-human-authorship-ai-generated-art-and-the-supreme-court/

16. Can AI Prompting Satisfy the Copyright Office's Human Authorship, https://www.tcamtoday.com/2026/can-ai-prompting-satisfy-the-copyright-offices-human-authorship-requirement/

17. latest from Allen v Perlmutter (AI authorship dispute) \- Reddit, https://www.reddit.com/r/COPYRIGHT/comments/1rhzi03/latest\_from\_allen\_v\_perlmutter\_ai\_authorship/

18. Copyright \- The Artificial Inventor Project, https://artificialinventor.com/copyright/

19. Thaler v. Vidal \- Fed Circuit Blog, https://fedcircuitblog.com/supreme-court/petitions/thaler-v-vidal/

20. Should Abandoned Applications Be Presumed Enabling? Supreme, https://patentlyo.com/patent/2025/12/abandoned-applications-presumed.html

21. Trade Secrecy Meets Generative AI, https://scholarship.kentlaw.iit.edu/cgi/viewcontent.cgi?article=4489\&context=cklawreview

22. DATA AS AN IP ASSET PART II INTA Data Protection Committee, https://www.inta.org/wp-content/uploads/public-files/advocacy/Data-as-an-IP-Asset-Part-2.pdf

23. EARNING TRADE SECRET S \- Cornell Law School, https://publications.lawschool.cornell.edu/lawreview/wp-content/uploads/sites/2/2024/12/Fishman-Varadarajan-final.pdf

24. The EU Data Act is Now Applying: Reflections on Data Access and, https://www.uni.lu/fdef-en/articles/the-eu-data-act-is-now-applying-reflections-on-data-access-and-intellectual-property/

25. AI and Trade Secrets: Why Trade Secret Law Is Your Best IP, https://www.robmelton.com/ai-and-trade-secrets-why-trade-secret-law-is-your-best-ip-strategy-for-ai-assets/

26. IP, AI, Digital Media & DMCA Policy | CocoJojo, https://cocojojo.com/legal/intellectual-property-ai-dmca-policy

27. The Promise and Perils of Enterprise Data as Trade Secrets, https://law.stanford.edu/wp-content/uploads/2026/01/Enterprise-Data-as-Trade-Secrets.pdf

28. Study in support of the evaluation of Directive 96/9/EC on the legal, https://technopolis-group.com/wp-content/uploads/2020/02/Study-in-Support-of-the-Evaluation-of-the-Database-Directive.pdf

29. 1\_EN\_ACT\_part1\_v8.docx \- EUR-Lex \- Europa.eu, https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52022PC0068

30. Database Protection & Sui Generis Rights \- Gevers, https://gevers.eu/blog/the-protection-of-modern-databases-under-the-sui-generis-database-right/

31. Data Act-Access to Data \- Open Future Foundation, https://openfuture.eu/wp-content/uploads/2022/02/Data-Act-Access-to-Data-1.pdf

32. The proposed EU Data Act and the sui generis database right, https://inquisitiveminds.bristows.com/post/102how2/the-proposed-eu-data-act-and-the-sui-generis-database-right

33. Why the Sui Generis Database Clause in the Data Act Is Counter, https://www.researchgate.net/publication/359904431\_Why\_the\_Sui\_Generis\_Database\_Clause\_in\_the\_Data\_Act\_Is\_Counter-Productive\_and\_How\_to\_Improve\_It

34. Overcoming Barriers to Text and Data Mining in the Era of ChatGPT, https://academic.oup.com/grurint/article/73/1/34/7286552

35. How and why the EU's Data Act will have an impact beyond, https://encompass-europe.com/comment/how-and-why-the-eus-data-act-will-have-an-impact-beyond-european-borders

36. Are AI models' weights protected databases? | Kluwer Copyright Blog, https://legalblogs.wolterskluwer.com/copyright-blog/are-ai-models-weights-protected-databases/

37. recent trends and developments in \- New York City Bar Association, https://www.nycbar.org/wp-content/uploads/2026/04/20221576-AnnualReportRecentTrendsAndDevelopmentsAI.pdf

38. Smart Contract Legal Enforceability: When Code Isn't Law, https://astraea.law/insights/smart-contract-legal-enforceability-code-isnt-law

39. Digital Assets and Web3: What Token Projects Need Before Going, https://www.daeryunlaw.com/es/practices/detail/digital-assets

40. Legislative Developments | March 9, 2026 \- CSC Blog, https://blog.cscglobal.com/legislative-developments-march-9-2026/

41. Securing the Rights of Cryptocreditors – Vol. 99, Issue 1 – HTML, https://ablj.org/securing-the-rights-of-cryptocreditors-vol-99-issue-1-html/

42. Robots, Markets, and the Value of Deal Lawyers Heather Hughes\*, https://jcl.law.uiowa.edu/sites/jcl.law.uiowa.edu/files/2024-06/Hughes\_Final.pdf