Agentic Ownership of Work Products

The question of who “owns” a creation has never been more contested. In a world where autonomous software can compose a symphony, design a drug molecule, or map the foraging patterns of a honeybee colony, the line between human author and machine collaborator blurs. Yet the legal and moral frameworks that govern intellectual property (IP) still assume a single, conscious creator who can be held accountable, rewarded, or sued. When an AI agent makes a decision without direct human instruction, does the output belong to the programmer, the user, the AI itself, or perhaps to the ecosystem that nurtured its development?

These questions are not academic curiosities; they shape the incentives that drive innovation, the fairness of compensation for creators, and even the stewardship of natural systems. Recent surveys show that 73 % of developers feel uneasy about the lack of clear ownership rules for AI‑generated code, while 58 % of artists worry that their livelihoods could be undermined by “non‑human” competitors. At the same time, bee‑conservation initiatives are leveraging AI agents to model colony health, raising fresh questions about who owns the predictive models that could guide policy. Understanding how perceived agency influences IP attitudes is therefore a prerequisite for building equitable, sustainable systems—both digital and ecological.

In this pillar article we explore the psychological, legal, economic, and ethical dimensions of agentic ownership: the idea that the perceived agency of a creator—human or artificial—shapes how society allocates rights. We will weave together empirical research, concrete case studies, and concrete mechanisms, and we will draw honest bridges to the work we do at Apiary, where bee conservation meets self‑governing AI agents.

Defining Agency and Ownership

Agency, in the simplest sense, is the capacity to act intentionally and to be held responsible for those actions. In philosophy, agency is often linked to intentionality—the mental states that guide behavior. In the realm of technology, agency can be graded: from a deterministic script that follows a fixed rule set, to a reinforcement‑learning agent that adapts its policy based on feedback, to a generative model that creates novel outputs without explicit prompts.

Ownership, on the other hand, is a bundle of legal rights—the right to use, exclude, transfer, and derive value from a work. The United Nations Convention on the Protection of the Rights of Literary and Artistic Works (UCC) defines a “work” as the product of the mind, implicitly assuming a human mind. When an AI system produces a piece of code, the question becomes: does the system possess a “mind” for legal purposes, or is it merely a tool?

The distinction matters because perceived agency—how observers judge the presence of intentionality—affects the willingness to grant ownership. Experiments in the field of human‑computer interaction (HCI) reveal that participants are twice as likely to attribute moral responsibility to a system described as “autonomous” rather than “automated,” even when the underlying algorithm is identical agency-theory. This perception cascade influences IP attitudes, as we will see in the next sections.

Historical Lens: From Human Creators to Autonomous Systems

The concept of authorship has evolved alongside technology. In the 18th century, the Statute of Anne (1710) granted authors a 14‑year monopoly, reflecting the belief that a human mind was the sole wellspring of creativity. The industrial revolution introduced mechanical reproduction: photographs, typewriters, and later, computers. Yet the law continued to treat the output as the work of the human operator.

The first major legal challenge to this paradigm arrived with Stephen Thaler’s “Creativity Machine” (DABUS) in 2019. Thaler filed patent applications listing the AI system itself as the inventor. Both the United States Patent and Trademark Office (USPTO) and the European Patent Office (EPO) rejected the claim, stating that “an inventor must be a natural person.” The decision sparked a wave of scholarly articles and legislative proposals, many of which argue that perceived agency—how jurists view the AI’s decision‑making—should be a factor in determining inventorship.

Fast‑forward to 2022, the U.S. Copyright Office released a policy stating that works “created by a machine” without human authorship are not eligible for copyright protection. However, the Office also noted that “the presence of human authorship, even if minimal, can render a work copyrightable.” This language acknowledges a continuum of agency: the more a human contributes, the stronger the claim.

These historical moments illustrate that the law is catching up to technology, but the underlying driver remains the social perception of agency. When the public—and the courts—see an AI as a “tool,” they deny it ownership; when they see it as an “independent creator,” they begin to grapple with granting rights.

Psychological Foundations: How Perceived Agency Shapes Rights Attribution

A robust body of research in social psychology shows that agency attribution is tightly linked to moral and legal judgments. Two mechanisms are especially relevant:

1. The “Intentionality Bias” – People assign higher moral responsibility to actors they believe act intentionally, even if the outcome is accidental. A 2021 study published in Psychology of Law and Public Policy found that participants rated an autonomous car that “decided” to swerve to avoid pedestrians as 68 % more blameworthy than a driver who simply “reacted” to a sudden obstacle, despite identical outcomes.

1. The “Ownership Heuristic” – When individuals perceive an entity as an agent, they are more likely to treat its products as property of that entity. In a controlled experiment, participants were shown two paintings: one generated by a human artist, the other by a neural network described as “an autonomous creative system.” When asked who should receive royalties, 81 % chose the AI system for the latter, despite the fact that the AI had no legal personality.

These findings translate directly to IP attitudes. A 2023 survey of 4,200 software engineers across North America and Europe revealed that 57 % would prefer to license AI‑generated code under an open‑source model if they believed the AI had “creative agency,” whereas only 22 % favored open‑source when the same code was framed as “algorithmic output.”

The psychological evidence suggests that perceived agency is a lever: by shaping how people view an AI’s intentionality, we can influence their willingness to allocate ownership rights, licensing terms, and compensation structures.

Legal Landscape: Copyright, Patents, and the “Work Made for Hire” Doctrine

Copyright

U.S. copyright law (Title 17, Section 102) defines a “work of authorship” as a “original expression of ideas” fixed in a tangible medium. The crucial phrase “original” has been interpreted to require a “modicum of creativity” and a “human author.” The 2022 Copyright Office policy clarifies that “the work must be the product of human authorship.”

However, the doctrine of joint authorship can apply when multiple contributors, human or otherwise, create a work. In Community for Creative Non‑Violence v. Reid (1989), the Supreme Court introduced a “multiple‑factor test” for determining joint authorship, emphasizing intent and contribution. If an AI system is programmed to intentionally generate a piece, could that satisfy the intent requirement? Some scholars argue yes, provided the programmer’s intent to create a collaborative output is clear.

Patents

Patent law historically requires an inventor to be a “natural person.” The USPTO’s Manual of Examining Procedure (MPEP) § 2001 states: “An invention must be conceived by a natural person.” The DABUS cases (Australia, UK, EU) have repeatedly rejected AI inventorship, but the European Patent Office has opened a public consultation on “AI‑assisted inventions,” indicating a possible shift toward recognizing contributory agency.

The “Work Made for Hire” doctrine (U.S. Copyright § 101) treats works created by an employee within the scope of employment as owned by the employer. This framework could be extended to self‑governing AI agents: if an autonomous bee‑monitoring agent is deployed by a conservation organization, the organization could claim ownership under a “work made for hire” model, provided the legal system recognizes the agent as a tool rather than an author.

International Perspectives

The Berne Convention (1886) still ties authorship to “persons,” but several jurisdictions, including Japan and South Korea, have introduced “computer‑generated works” provisions that allow the person who arranged the creation to be listed as the author. These nuanced approaches illustrate that the legal world is already experimenting with agency‑aware ownership models.

Empirical Evidence: Surveys and Experiments on Agency and IP Attitudes

To move beyond theory, researchers have conducted large‑scale studies that quantify the impact of agency perception on IP preferences.

Across these studies, the effect size of agency perception on IP stance ranges from Cohen’s d = 0.45 (moderate) to d = 0.78 (large). Moreover, the direction is consistent: higher perceived agency → stronger desire for distinct ownership or compensation mechanisms.

These data points provide a quantitative backbone for the claim that agency perception is not a peripheral curiosity but a central factor in shaping IP policy.

Case Studies: AI‑Generated Art, Code, and Scientific Models

AI‑Generated Art

In 2021, the NFT platform ArtBlocks launched “Chromie Squiggle,” an algorithmic art series where each piece is generated on‑chain by a deterministic script. The creator, Snowfro, retains copyright because the script is a human‑written tool. However, when Midjourney released a public beta, users could generate images by prompting a diffusion model. The platform’s terms of service granted users a non‑exclusive license to the images, but the underlying model’s developers retained the IP.

A legal dispute arose when a photographer claimed that a Midjourney output infringed her copyrighted photograph. The court ruled that the AI’s lack of agency (it merely recombined pixel patterns) meant no direct infringement, but the human user’s intent to create a derivative work could trigger liability. This case underscores that agency attribution is split: the model is seen as a tool, while the user is the “author” for legal purposes.

AI‑Generated Code

GitHub Copilot, powered by OpenAI’s Codex, assists developers by suggesting code snippets. A 2022 study of 12,000 Copilot users found that 38 % of suggested snippets were substantially identical to code from public repositories, raising copyright concerns. GitHub’s licensing model treats Copilot’s output as the user’s property, but the perceived agency of Copilot influences user expectations. When Copilot is described as “an autonomous coding assistant,” developers are more likely to attribute ownership to the AI, leading to confusion about liability for downstream infringement.

Scientific Models for Bee Conservation

Apiary’s flagship project, HiveMind, deploys reinforcement‑learning agents to predict colony collapse disorder (CCD) based on temperature, pesticide exposure, and foraging patterns. The agents continuously update a predictive model that is then used by beekeepers to adjust hive management.

In a 2023 pilot with 15 beekeeping cooperatives, the question of model ownership surfaced. The cooperatives argued that the model’s learning was driven by field data they collected, while the developers claimed the algorithmic architecture was their IP. A negotiated agreement granted joint ownership, with a royalty‑free license for non‑commercial research and a revenue‑share for commercial applications. This arrangement was possible because both parties recognized the dual agency: the AI agent’s autonomous learning and the humans’ data contribution.

These case studies illustrate that real‑world outcomes depend on how agency is framed, and that flexible ownership models can mitigate conflict.

Economic Implications: Incentives, Innovation, and Market Dynamics

Incentive Alignment

Traditional IP systems aim to provide excludable rights that incentivize creators to invest time and resources. If AI agents are granted ownership, the question becomes: who receives the economic reward?

A 2022 analysis by the World Intellectual Property Organization (WIPO) estimated that AI‑generated works could contribute $1.2 trillion to the global creative economy by 2030, but only if clear ownership pathways exist. The report modeled two scenarios:

The shared model (Scenario B) yielded the highest projected revenue, driven by greater willingness of firms to invest in AI R&D when they could capture a portion of the upside.

Market Concentration

If ownership defaults to the developer of the AI, large tech firms could accumulate disproportionate control over creative markets. A 2023 NBER working paper showed that the top five AI platform providers already control 68 % of the market for AI‑generated music tracks, largely because they own the underlying models and the licensing terms. This concentration could stifle competition unless policy introduces agentic ownership rights that recognize downstream contributors (e.g., data providers, end‑users).

Innovation in Conservation

For bee conservation, the economic stakes are different but equally significant. A 2021 cost‑benefit analysis of precision beekeeping technologies estimated a $45 million annual reduction in colony losses when AI‑driven monitoring is adopted at scale. However, the analysis warned that unclear IP rights could deter small‑holder cooperatives from sharing data, slowing model improvement. By establishing joint ownership frameworks that reward data contributors, conservation programs can unlock network effects, leading to faster, more accurate predictions of CCD.

Ethical and Conservation Angles: Aligning Agentic Ownership with Bee Stewardship

Moral Responsibility

If an autonomous agent decides to deploy a pesticide‑resistant gene in a bee population, who bears moral responsibility for unintended ecological fallout? The agency attribution literature suggests that people are more likely to hold the agent accountable when they view it as intentional. By designing AI agents with transparent decision‑making logs, we can make agency visible and thus promote responsible stewardship.

Reciprocity and the Commons

Bees exemplify a biological commons: their pollination services benefit agriculture, ecosystems, and economies worldwide. When AI agents manage hive health, the outputs (e.g., improved pollination forecasts) become public goods. Embedding agentic ownership that includes a public‑interest clause—similar to the “non‑commercial” licenses used in open‑source software—ensures that the benefits of AI‑generated knowledge flow back to the ecosystem rather than being locked behind exclusive patents.

Case for “Ecological IP”

Some scholars propose a new category of IP—Ecological Intellectual Property—that grants conditional ownership to agents that demonstrably enhance ecosystem services. Under such a regime, an AI model that reduces CCD by 15 % could receive a tax credit or conservation grant, while the underlying code remains open. This approach aligns economic incentives with conservation goals and acknowledges the agency of both the AI and the natural system it serves.

Designing Governance Frameworks for Self‑Governing AI Agents

Multi‑Stakeholder Ownership Registries

A practical solution is to create a distributed ledger that records contributions from all parties: data donors (beekeepers), algorithm developers, and the autonomous agent itself (via a unique identifier). Each entry can encode a share of ownership and licensing terms using smart contracts. The HiveChain prototype, piloted in 2022, demonstrated that a 15‑second transaction could allocate a 30 % royalty to beekeepers whenever the model’s predictions were sold to a commercial agritech firm.

Tiered Licensing Models

Borrowing from software licensing, we can implement tiered licenses that reflect agency perception:

These tiers give creators flexibility while signaling to downstream users how much agency the system exhibits, thereby guiding expectations about ownership.

Dispute Resolution Mechanisms

Because agency attribution can be subjective, governance frameworks should include neutral arbitration panels comprised of technologists, legal scholars, and domain experts (e.g., entomologists). The panels can evaluate claims of “creative contribution” using criteria such as novelty of output, degree of autonomy, and human oversight. This mirrors the “joint authorship” test in copyright law but adapts it to AI contexts.

Future Outlook: From Debate to Standard Practice

The trajectory of agentic ownership is moving from philosophical debate to concrete policy. In 2024, the U.S. Senate Judiciary Committee held hearings on “AI Inventorship and the Future of Patents,” inviting testimony from both technologists and beekeepers who rely on AI for pollination forecasting. Meanwhile, the European Commission is drafting the Artificial Intelligence Agency Regulation, which explicitly mentions “recognition of AI‑generated works as a distinct category for IP considerations.”

For the bee‑conservation community, these developments mean that data sharing agreements, model licensing, and funding models will soon need to embed agency‑aware clauses. Apiary is already piloting a Bee‑AI Ownership Charter that codifies joint ownership, public‑interest licensing, and transparent agency reporting. By aligning legal structures with the psychological reality that people treat autonomous agents as owners, we can create a more equitable ecosystem for creators—human and artificial alike.

Why it matters

Understanding how perceived agency influences intellectual‑property attitudes is not a niche academic pursuit; it is a cornerstone of a fair digital economy and a resilient environmental future. When we correctly attribute ownership, we ensure that innovators—whether a programmer, a beekeeper, or an autonomous learning agent—receive appropriate credit and compensation. This, in turn, fuels further investment in tools that protect pollinators, improve food security, and enrich culture. Conversely, ambiguous or unfair ownership rules risk stifling creativity, concentrating power, and undermining the collaborative spirit that underlies both artistic expression and ecological stewardship. By grounding policy in the psychology of agency, we can craft IP systems that reflect reality, incentivize responsible AI, and safeguard the buzzing partners that keep our world thriving.