We are pleased to present the latest edition of CRA’s IP Literature Watch. This issue contains pieces on antitrust & IP, licensing, litigation, innovation, law and policy, copyright, and trade.
This newsletter contains an overview of recent publications concerning intellectual property issues. The abstracts included below are as written by the author(s) and are unedited.
IP & Antitrust
Antitrust Theory Meets the Real World: Empirical Shortfalls of the New Conventional Wisdom on Competition Policy
Jonathan Barnett (USC Gould School of Law)
USC CLASS Research Paper No. 2613
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6636318
For at least half a century, antitrust enforcement and jurisprudence have extensively adopted economic methods and substantially replaced structuralist approaches with a focus on consumer welfare, balancing tests, and error-cost principles. New and increasingly popular schools of thought seek to refashion competition law by discounting economic methods, expanding policy objectives to encompass redistributive and other non-economic concerns, adopting categorical prohibitions, and relaxing error-cost constraints. The rise of this interventionist approach has coincided with a “transmission pipeline” through which emergent theories of anticompetitive harm, based on theoretical models, preliminary findings, or narrow evidence developed in the scholarly literature, are adopted expansively by advocates and regulators who pursue enforcement actions with significant consequences. Examining theories of patent holdup, killer acquisitions, kill zones, and default effects, this paper shows that competition agencies in the US, EU, and other major jurisdictions have relied on at least four liability theories with contestable or weak empirical or theoretical foundations. This “rush to intervene” endangers rule-of-law values and, without meaningful balancing tests, can yield market outcomes that harm the consumers and entrants that antitrust law seeks to protect. The alternative trajectory of common ownership theory, which elicited robust scholarly critique and largely failed to translate into enforcement action, suggests that, especially in industries without political salience, the transmission mechanism from academia to policy sometimes functions well to filter out insufficiently demonstrated theories of anticompetitive harm.
IP & Licensing
Controlling Market Power in ASEAN: A Comparative Institutional Law and Economics Analysis of the Intellectual Property-Competition Law Interface
Robert Ian McEwin (National University of Singapore Faculty of Law; Independent)
submitted to Editors of Oxford Handbook
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6993219
Much of the scholarship concerning the intellectual property–competition law interface remains heavily doctrinal. Analysis commonly focuses upon compulsory licensing, refusals to deal, standard-essential patents (SEPs), patent settlements, technology transfer agreements and platform access obligations. Such work has generated valuable insights. However, it frequently assumes institutional environments like those of advanced jurisdictions, particularly those of the United States and the European Union.
This chapter argues that the interaction between intellectual property and competition law cannot be understood adequately through doctrinal analysis alone. The appropriate balance between intellectual property protection and competition law intervention depends critically upon institutional capacity, development strategy, market structure and political economy. Legal rules do not operate in the abstract. Their economic significance depends upon the institutions that interpret, administer and enforce them.
IP & Litigation
Patent Litigation Finance and the Venture Capital Myth
Sean Thompson (Parabellum Capital)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6797642
A recent Sedona Conference Journal article contends that patent litigation funders employ a “venture capital model,” financing high volumes of low-merit cases on the theory that occasional outsized winners will offset losses of 50 to 70 percent. The authors use that descriptive claim about how the asset class operates to motivate disclosure and fee-shifting reforms.
This Article shows that the VC model is not a strategy any rational patent litigation funder would pursue, and the available evidence indicates that none in fact does. The analysis proceeds along three lines.
The evidentiary basis for the VC model claim is one enforcement campaign and two individual cases. None produced the collectible outlier recovery the theory requires. The strategy also fails as a matter of fund mathematics. A portfolio carrying 50 to 70 percent losses requires winners large enough to cover the lost capital and clear the return hurdles that allow a manager to raise a successor fund. Patent damages are bounded by statute and doctrine, compressed on appeal, and realized over decade-long timelines. Recent empirical work shows damages awards above $500 million have a 0 percent survival rate on appeal. The patent system does not produce the collectible tail outcomes the VC model requires.
The publicly reported returns of Burford Capital and Omni Bridgeway, the two largest publicly traded funders, show loss rates of 15 and 22 percent of deployed capital and moderate, distributed winners rather than rare lottery payoffs. That is the opposite of a VC-style return profile.
Patent litigation finance is viable because competent practitioners do the opposite of what the VC model describes. The asset class rewards selectivity, not variance. Reform proposals grounded in an inaccurate description of the market produce unsound policy.
Who Creates and Who Captures? Generative AI, Copyright Infringement, and Optimal Policy
Hongwei Kou (Tsinghua University)
Ke Rong (Institute of Economics, School of Social Sciences, Tsinghua University)
Danxia Xie (Tsinghua University – Institute of Economics)
Buyuan Yang (Central University of Finance and Economics (CUFE) – School of International Trade and Economics)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6850760
Generative AI gives creative works a dual role as final goods and training inputs. This dual role creates an appropriation problem: copyrighted knowledge strengthens AI and supports content production through non-rival AI use, but infringing use of such knowledge erodes returns to original creation. We develop a general equilibrium model with an upstream AI sector, an original sector, and a derivative sector. In the decentralized equilibrium, labor is under allocated to original production and overallocated to AI production because private agents do not internalize the cross-sector effects of non-rival AI use and the knowledge spillovers embedded in AI training; the derivative sector’s AI use creates no separate distortion. We examine two governance instruments: digital taxation and copyright litigation. Optimal taxation is asymmetric: the derivative sector is always taxed, while the original and AI sectors may be taxed or subsidized depending on the strength of appropriation and the value of AI as a nonrival input. Copyright litigation has an inverted-U welfare effect: moderate enforcement reallocates returns toward original creators through a beneficial compensation effect, whereas excessive enforcement contracts AI supply and downstream output through an adverse production effect. The optimal policy depends on the relative costs of tax administration and legal enforcement.
IP & Innovation
Beyond Conception: AI, the America Invents Act, and the Temporal Anchor of Invention
David S. Olson (Boston College Law School)
Boston College Law School Legal Studies Research Paper No. 681
Brooklyn Law Review, Vol. 92 (Forthcoming)
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6951678
Artificial intelligence systems increasingly generate technical solutions before any human has anticipated, formulated, or understood them. Under current patent doctrine, these inventions may be unpatentable—not because they fail requirements of novelty, utility, or disclosure, but because they lack a psychologically framed moment of human conception. This Article argues that the continued centrality of conception is both historically contingent and doctrinally unnecessary. Conception arose as a priority-allocating device under the first-to-invent regime. The America Invents Act (AIA) eliminated that regime, replacing it with a first-inventor-to-file system in which priority turns on filing date. Yet conception persists as the definitional anchor of invention, retained by doctrinal inertia rather than functional necessity. Every substantive concern that conception might plausibly serve (screening for operability, ensuring disclosure, preventing trivial patenting) is already performed more directly by existing requirements of utility, enablement, written description, novelty, and obviousness. This Article proposes relocating the temporal anchor of invention from mental formulation to demonstrable technological achievement through actual or constructive reduction to practice. This realignment harmonizes patent doctrine with the AIA’s statutory structure, accommodates computational modes of innovation without inventorship metaphysics, and preserves every existing safeguard against speculative or overbroad patenting.
Quality Adjustment in Industry Deflators Strengthens Estimated Innovation–Productivity Relationships
Enghin Atalay (Federal Reserve Banks – Federal Reserve Bank of Philadelphia)
Ali Hortaçsu (University of Chicago; National Bureau of Economic Research (NBER))
Nicole Kimmel (Federal Reserve Banks – Federal Reserve Bank of Philadelphia)
Chad Syverson (University of Chicago – Booth School of Business; National Bureau of Economic Research (NBER))
FRB of Philadelphia Working Paper No. 26-22
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6851112
How do investments in innovation translate into future productivity growth? Empirically answering this question is challenging. R&D spending is an observed input into the innovation process, but mapping it to productivity growth requires assumptions about the depreciation of R&D capital, gestation lags, and how well such expenditures capture true innovative effort (Hall, 2007). Patents, an alternative measure, capture successful innovations but vary widely in novelty (Kelly et al., 2021) and economic value (Kogan et al., 2017). Firms may forgo patenting to preserve secrecy, while others patent strategically to protect existing products even when their underlying innovations are marginal.
Key Patents and Economic Performance: Patent Quality, Technological Innovation, and Growth in Knowledge-Based Economies
Maximilian Wurster (Independent)
Detlef Hellenkamp (Duale Hochschule Baden-Württemberg)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6809941
In knowledge-based economies, technological innovation capability, intangible assets, and intellectual property are becoming increasingly important for long-term economic performance. Patent data are widely used as indicators of innovation; however, simple patent counts do not adequately capture the quality, technological relevance, or international market coverage of individual patents. This paper therefore examines the extent to which key patents can serve as a quality-oriented indicator of technological innovation capability for analyzing economic performance. Key patents are understood as particularly high-quality patents in relevant key technologies, whose patent strength derives from technological relevance and international market coverage.
The paper combines a conceptual analysis of the key patent approach with literature from growth and innovation economics, as well as with an exploratory empirical plausibility assessment based on a global patent dataset on world-class patents in future technologies (Breitinger et al., 2020). Economic performance is operationalized empirically through per capita income. The empirical evidence is assessed from two complementary perspectives: an international cross-section of high-income economies and a dynamic analysis of the United States, which held a particularly prominent position in high-quality patenting across key future technologies during the period under observation. In both perspectives, the findings point to a positive association between key patents and per capita income. They should not, however, be interpreted as evidence of causality, but as empirically grounded support for the plausibility of this relationship.
The contribution of this paper lies in positioning key patents as a differentiated innovation indicator that takes patent quality, technological relevance, and international reach more fully into account than mere patent volumes. The key patent approach thus offers a scientifically grounded perspective on innovation capability and economic performance in knowledge-based economies.
Peer Pay Inequality and Innovation: Evidence from Salary History Bans
Simi Kedia (Rutgers Business School)
Chunka Tai (University of Connecticut)
Lingling Wang (University of Connecticut – Department of Finance)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6788360
We distinguish horizontal, or peer-level, pay dispersion from vertical pay dispersion and examine its impact on innovation. Using the staggered adoption of state-level salary history bans, we show that the laws significantly reduce horizontal pay dispersion while leaving vertical pay dispersion largely unchanged. Treated firms experience an 11-16% increase in patent value, with stronger effects where peer-pay inequality is more pronounced and innovation depends more heavily on employee effort and collaboration. Consistently, treated firms experience lower employee turnover and greater team-level collaboration and productivity. Our findings suggest that mitigating pay inequality among peers improves innovation.
IP Law & Policy
As Humans Step Aside: Intellectual Property, Artificial Intelligence and Drug Development
Frederick M. Abbott (Florida State University – College of Law)
Research Handbook in Intellectual Property and Health (Ana Rutschman ed.) under contract with Elgar Publishing, forthcoming 2026/7
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6966638
This chapter addresses the increasingly important question whether artificial intelligence (AI) may fulfill the dreams of its proponents and begin to solve the real problems of resource scarcity, particularly of affordable new drugs. Leaders in the AI community have floated the idea that artificial general intelligence (AGI) will change the world to one where individuals will contemplate what to do when freed from the demands of labor because resource utilization will be maximized. We will live in an era of abundance. The same leaders have suggested that AI and/or AGI will tackle human biology and potentially find cures for all diseases.
It seems almost certain that AI and/or AGI will be successful in establishing a much better understanding of human biology and ways to modify it to promote health. It is unlikely that intellectual property rules will play a material role in determining the path forward. Yet IP rules may help to determine the speed and scale at which the public has access to the benefits of the new technologies, and as always with IP there are risks of under-protection and overprotection that need to be balanced. As has been the case in previous iterations of challenge to conventional IP rules, there will be choices between stepwise adaptation of existing rules and the adoption of sui generis rules to address the changed environment.
We proceed in two steps. The first is to consider the issues raised by introducing AI generated inventions into conventional IP systems, most prominently patent systems. The second is to contemplate the possibility of new paradigms that are developed on a fresh slate based on evolving requirements, and not looking back on the conventional modalities.
Narrative Amplification in Drug Patent Scholarship
David S. Olson (Boston College Law School)
Boston College Law School Legal Studies Research Paper No. 680
American University Law Review (Forthcoming)
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6951618
A majority of legal scholarship argues that pharmaceutical companies abuse patent law to extend exclusive drug rights, in order to obtain high drug prices for much longer periods than Congress intended. A counter-narrative argues that the practices being labeled as abuses are not necessarily the primary drivers of excessive drug prices and are instead functioning as intended: to encourage the ongoing improvement of lifesaving medicines. This counter-narrative argues that suggested reforms will decrease drug innovation and harm patients. This important debate raises the questions: What are the facts and which side is right? This Article does not answer these questions. This Article asks a critical prior question: How faithfully does the literature engaged in this debate transmit what empirical studies actually show?
Using an original citation-network analysis of 84 articles, this Article analyzes narrative amplification in legal scholarship by tracing empirical claims to their root studies and coding both transmission fidelity and engagement depth. Three patterns emerge. First, between 53% and 100% of the central inferential constraints tracked in the root studies are not transmitted in downstream articles. Second, downstream articles selectively engage with empirical studies. Third, counterevidence is largely absent from articles that attribute excessive drug prices to the patent system; only 5% of such articles cite any counter-narrative empirical study, even when they cite counter-narrative scholars’ non-empirical work. This Article offers a replicable method for identifying when limited empirical findings transform into broader accepted truths, and explores why legal scholarship is particularly susceptible to narrative amplification.
Copyright Law
The Washington Effect of the US Copyright Act
Peter K. Yu (Texas A&M University School of Law)
Journal of the Copyright Society, Vol. 73, 2026, Forthcoming
Texas A&M University School of Law Legal Studies Research Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6985124
On October 19, 1976, the Copyright Act of 1976 will celebrate its fiftieth anniversary. Published as part of a commemorative special issue on the internationalization of U.S. copyright law, this article examines the cross-jurisdictional influence, or “Washington effect,” of the U.S. Copyright Act, including both the 1976 Copyright Act and subsequent legislative enactments. It begins by examining the United States’s transformation from a country offering limited protection to nonresident foreign authors to one championing high levels of copyright protection around the world. The article then provides two leading illustrations of U.S. copyright law transplants—namely, fair use and the notice-and-takedown procedure. It concludes by drawing five key takeaways from the process of transplanting U.S. copyright standards around the world. The discussion reveals the extent and limitations of the Washington effect, or cross-jurisdictional influence, of the U.S. Copyright Act.
A Meditation on Fixation as a Line Between Federal and State Copyright
Timothy McFarlin (Samford University – Cumberland School of Law)
73 J. Copyright Soc’y (forthcoming 2027)
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6963363
In 1976, Congress decided to change the line between federal and state copyright in works of authorship. Instead of when such works were published, or registered before publication, the new divide became fixation. Previously, state copyrights existed in many unpublished works of authorship. But ever since the 1976 Act, state copyrights can only exist if the works are “not fixed,” e.g., if they’ve been orally communicated but have yet to be written down by (or under the authority of) their author in a stable form. This Essay is a meditation on that dividing line and what it may yet mean for U.S. copyright, including how it might lead to another federal-state divide: one between human- and AI-generated works.
Copyright’s Public Domains: The Limits on AI Appropriation
Graham Greenleaf (Macquarie University – Macquarie Law School (Sydney, Australia))
David F. Lindsay (UTS: Law)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6840700
The spectacular rise to commercial and intellectual prominence of artificial intelligence (AI) since 2022, and in particular the predominant role of generative AI and its use of large language models (LLMs), has given rise to many legal and policy problems. Four main types of problems for copyright law are sketched, in each of which one significant legal question raised is whether the use of the content is in the public domain, or is it an infringement of copyright? Can the developer of the training set, or the deployer of the AI system, or the end-user of the AI system, claim that the use they have made of the AI system’s content does not involve an infringement of copyright but is a use of that content which is in the public domain?
Those building AI systems are faced with three alternatives. They can negotiate to obtain the consent (through private licenses) of the owners of copyright in the works intended to be used. Or they can ignore whether or not consent may be necessary, on the basis that they ‘can get away with it’ anyway. Or they can attempt to justify the use they wish to make of the content on the basis that it does not require owner consent because it is in the public domain. This last approach is the subject of this article, which aims to be comprehensive and precise about which aspect(s) of the public domain can be used to support claims that consent is not legally necessary.
Our previous work has identified fifteen aspects of the copyright public domain, the aggregate of which is the copyright public domain in a particular jurisdiction. We consider each of those fifteen categories, with a focus on Australian law, explaining first how the category is consistent with our overall definition of the public domain (that is, ‘the public’s ability to use content on equal terms without seeking permission’). Then the relationship of each category to international copyright law is stated, with an emphasis on how much (if any) expansion of that category in national laws is consistent with current international copyright law. We give brief examples of how each category is reflected in national laws, particularly in Australia. Finally, we consider possible ‘opportunities’, meaning how this aspect of the public domain has been or could be used or expanded (by legislation or case law) to assist the development of AI systems.
This article concludes that developing solutions for the substantial challenges posed by generative AI can be assisted by analyzing the actual and potential extent to which all of the existing public domain categories apply, or could reasonably be developed to apply, to access and use for developing AI systems. Instead of a ‘magic bullet’ to be found in only one category, the best solution might come from a combination of various public domain elements.
From our survey we find that nine of the fifteen public domain categories offer some possibilities for supporting AI development, but in most categories these possibilities are slight. Two most clearly permit use of a substantial amount of content without the need for permission: the substantial number of works in which the copyright term has expired (category 5); and material available for use under neutral voluntary licenses (category 14), such as CC licenses. In two further categories access to significant amounts of public domain content is complicated due to legal uncertainties and practical obstacles: fair use exceptions, at least in the US and the EU (but not at present in Australia); and the tenet of copyright law that mere facts or ideas (category 10) can be freely used, which might seem to hold considerable potential for lawful AI training. Other public domain categories, as they currently exist, are of only theoretical value.
The question then turns to the extent to which there is potential for developing existing public domain categories, while appropriately balancing the interests of authors and owners, on the one hand, and AI developers, on the other. There is more scope for national jurisdictions to reform the public domain categories than is commonly thought. For example, works expressly excluded from copyright protection (category 3), could take advantage of the two permissible optional exclusions, namely laws and other official texts, and political and legal speeches, and ‘news of the day’. This leaves the two public domain categories that have been a focus of debates about policy responses to the rise of generative AI: ‘free use’ exceptions (category 12) and neutral compulsory licensing (category 13). We can see some scope for a very nuanced TDM exception for some AI training which is in the public interest, and for some conventional compulsory licenses or extended collective licenses (ECLs). Reliance on the status quo, and market-based approaches are unlikely to be sufficient.
Copyright and Artificial Intelligence: The Importance of Memorization and Attribution
Christian Koboldt (DotEcon)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6814498
Generative AI raises copyright questions that earlier copying technologies did not. This paper distinguishes two markets in which AI and copyright interact – the market for training data access and the market for reproduced outputs – and argues that they raise distinct welfare problems requiring distinct policy responses. Drawing on the economics of information goods, the technical literature on memorization in large language models, and recent legal developments in Germany, the United Kingdom and the United States, we build on Gans (2024) to propose a capability-based levy: payments to rights holders calibrated to the measurable memorization rate of deployed models. This reflects potential harm in the reproduced outputs market rather than identified training-data use. We argue that the distribution problem inherent in any such mechanism, together with the non-pecuniary harm to creators from unattributed use of their work, can be addressed by mandating or incentivizing source-attribution capability at the model-architecture level. Relative to training-data-based schemes, the capability-based approach better aligns the payment obligation with actual harm, creates incentives for AI developers to reduce memorization, and is less susceptible to territorial arbitrage.
IP & Trade
Pharmaceutical Intellectual Property, Data Exclusivity and Drug Prices – A Law and Economics Analysis of Access To Medicines in The EU And Pakistan
Ans Iqbal (University of Szeged – Faculty of Law, Graduate School)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6804279
The current article delves into the impact of pharmaceutical intellectual property (IP) laws and regulatory exclusivities on access to medicines in the European Union (EU) and Pakistan within a law and economics framework. The study is based on a doctrinal review and analysis of patents, data exclusivity, and pricing policy, supplemented by evidence from European policy and industry reports and empirical findings from Pakistan’s health research, including those conducted by Aga Khan University (AKU). It believes that the EU’s patent thicket and regulatory monopoly give clear incentives to innovate but lead to clear disparities across countries in the supply and price of new medicines. The TRIPS-compliant patent regime in Pakistan is subject to similar, albeit distinct, challenges of access but comes with a fiscal space shortage, institutional capacity limitations, and strong dependence on generics. The article takes the contrast between the EU and Pakistan as an example to analyze the interplay between global IP norms and domestic regulatory options and their impact on static and dynamic efficiency for pharmaceutical markets. The document ends with suggestions for reforms in both systems and a research program for a larger thesis on European and international drug regulation.
Other Topics
The Protection of Human Voice in the Age of AI
Tamika Mansell (Macquarie University – Macquarie Law School (Sydney, Australia))
Rita Matulionyte (Macquarie Law School)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6807698
This paper examines the effectiveness of Australian copyright law and the passing off doctrine for protecting the human voice against unauthorized AI replication and use. It compares these frameworks with current US legal approaches to voice protection, namely copyright, the right of publicity and emerging name, image, voice and likeness legislation. This analysis reveals the inadequacy of existing Australian law to safeguard vocal identity in the age of AI and proposes a legislative right of voice as a potential remedy. By providing a much-needed Australian perspective on this emerging area, this paper holds potential to generate valuable discourse and offer guidance to policy makers addressing the challenges posed by AI voice technology.
Greening the Deal: Can Mergers Redirect Innovation?
Melissa Newham (ETH Zürich – CER-ETH – Center of Economic Research at ETH Zurich; KU Leuven)
David Jaggi (Zurich University of Applied Sciences)
Jan-Alexander Posth (ZHAW School of Management and Law)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6841578
This paper examines how mergers and acquisitions (M&As) affect the direction of corporate innovation. Using data on U.S. M&A transactions from 1988-2015 matched with detailed patent data, we construct novel measures of changes in the technological orientation of acquirers based on patent text embeddings. We estimate dynamic treatment effects in a staggered difference-in-differences framework that compares merging firms to matched controls over a 17-year period, centered on the deal announcement year. We find a large and statistically significant shift in the acquirer’s innovation direction toward the target’s knowledge base post-deal. This positive effect is driven by deals between acquirers and targets that operate in technologically distinct domains. Using a subsample of deals involving targets that hold clean patents, we find that “dirty” acquirers – firms whose patent portfolios are concentrated in dirty technologies – show particularly strong post-deal shifts toward their target’s clean technologies, accompanied by increased citations to the target’s clean patents and increased clean patent filings by the acquirer. Overall, our results point to the potential for M&A to reshape innovation trajectories and mitigate against path dependencies.
Data Labelling at the Legal Frontier: Liability, Consent, Security Obligations, and the Emerging Regulatory Architecture of the AI Annotation Supply Chain Interdisciplinary Analysis of Law, Technology & Policy
Divya Pandey (OWOW Talents Inc)
Gangesh Pathak (OWOW Talents Inc)
Shivani Manchanda (San Francisco Bay University)
Nishant Sonkar (Cisco Systems)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6871038
Data labelling-the human annotation of raw data to train artificial intelligence systems-sits at the critical nexus of technology law, intellectual property, privacy regulation, labor rights, and cybersecurity. As the global AI training-data market expands rapidly-on most analyst estimates growing at over 20 per cent annually toward the low tens of billions of dollars by the mid-2030s-the legal obligations surrounding how data is collected, who labels it, and how it is secured have become urgent priorities for developers, regulators, and practitioners alike. This paper synthesizes the latest academic literature, regulatory frameworks, landmark litigation, and industry data across three interlocking pillars: (1) Liability-who bears legal responsibility when annotated training data causes harm, infringes copyright, or produces biased AI outputs; (2) Consent-under what conditions personal or creative data may be lawfully collected, shared, and annotated, and what rights data subjects and workers retain; and (3) Security Obligations-the cybersecurity and data-governance duties that apply across the annotation supply chain. It further analyses twelve major categories of security risk specific to the data-labelling ecosystem, from data poisoning and adversarial label manipulation to geopolitical cross-border transfer risks. The analysis demonstrates that three converging forces are reshaping the legal landscape: regulatory convergence through the EU AI Act, revised Product Liability Directive, and a wave of US state AI laws; mounting litigation pressure across copyright, product liability, and labor law; and progressive recognition by international bodies of annotation workers’ data-protection and labor rights. Organizations that invest now in data provenance documentation, worker rights compliance, and zero-trust security architecture will be better positioned to navigate the regulatory and litigation risks that will intensify through 2026 and beyond.
A Comprehensive Geocoded Dataset of All Chinese Patent Applicants
Dongbo Shi (Beijing Institute of Mathematical Sciences and Applications)
Fan Zhang (Digital Economy Lab, BIMSA)
Working Paper
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6818519
This paper presents the construction of an open-source dataset of geocoded applicant addresses for invention patent applications filed with the China National Intellectual Property Administration (CNIPA) from 1985 to 2024. We first impute missing addresses for 1,737,979 patent-applicant instances, all of which correspond to non-first applicants whose addresses are not disclosed in the raw CNIPA data. We then geocode all addresses and successfully geocode address for 19,994,120 patent-applicant instances, covering 98.72% of all patent-applicant instances, with 85.33% geocoded to the district/county level or finer. The resulting dataset improves the spatial completeness of Chinese innovation data and provides a high-resolution resource for studying geographical patterns of innovation production, innovation collaboration, and knowledge spillovers.
