January 01 2024 The Work of Copyright Law in the Age of Generative AI Kate Crawford, Kate Crawford Kate Crawford is a Research Professor at USC Annenberg, a Senior Principal Researcher at MSR-NYC, and founder of the Knowing Machines Project. Her latest book, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence (Yale University Press) won the Sally Hacker Prize (2022). Search for other works by this author on: This Site Google Scholar Jason Schultz Jason Schultz Jason Schultz is a Professor of Clinical Law at NYU School of Law, Director of NYU's Technology Law & Policy Clinic, and Codirector of the Engelberg Center on Innovation Law & Policy. Search for other works by this author on: This Site Google Scholar Author and Article Information Kate Crawford Kate Crawford is a Research Professor at USC Annenberg, a Senior Principal Researcher at MSR-NYC, and founder of the Knowing Machines Project. Her latest book, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence (Yale University Press) won the Sally Hacker Prize (2022). Jason Schultz Jason Schultz is a Professor of Clinical Law at NYU School of Law, Director of NYU's Technology Law & Policy Clinic, and Codirector of the Engelberg Center on Innovation Law & Policy. Online ISSN: 1536-0105 Print ISSN: 1526-3819 © 2024 Kate Crawford and Jason Schultz2024Kate Crawford and Jason Schultz Grey Room (2024) (94): 56–62. https://doi.org/10.1162/grey_a_00389 Cite Icon Cite Permissions Share Icon Share Facebook Twitter LinkedIn Email Views Icon Views Article contents Figures & tables Video Audio Supplementary Data Peer Review Search Site Citation Kate Crawford, Jason Schultz; The Work of Copyright Law in the Age of Generative AI. Grey Room 2024; (94): 56–62. doi: https://doi.org/10.1162/grey_a_00389 Download citation file: Ris (Zotero) Reference Manager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex toolbar search Search Dropdown Menu toolbar search search input Search input auto suggest filter your search All ContentAll JournalsGrey Room Search Advanced Search This content is only available as a PDF. © 2024 Kate Crawford and Jason Schultz2024Kate Crawford and Jason Schultz Article PDF first page preview Close Modal You do not currently have access to this content.

of Law provides a unique environment where scholars can examine the key drivers of innovation as well as the law and

Have you ever noticed that you can’t really “buy” an ebook? Sure, when you click that “Buy Now” button on your ereader, tablet, or phone, it feels like a complete, seamless transaction. But the minute you try to treat your ebook like a physical book – say by sharing it with a friend, selling it to someone else, donating it to a school library, or sometimes even reading it offline, reality sets in. You can’t do any of those things.With most ebooks, even if you think you “own” them, the publisher or platform you bought them from will say otherwise. Publishers and platforms insist that you only buy a license to access the books, not the rights to do anything else with them. And because platforms like Amazon and Apple control most of the technology we use to read ebooks, their opinion often dictates the reality of the ebook ecosystem. Beyond controlling books, these platforms can also do several things no physical bookseller has ever had the power to do. They can track your reading habits, stop you from reselling or lending a book, change the book’s content, and delete it from your digital library altogether – even after you’ve bought it. This doesn’t happen in the print book market, where you can still feel confident that when you buy a book, it’s yours to share, sell, or simply read without it being tracked or censored.Something happened when we shifted to digital formats that created a loss of rights for readers. Pulling back the curtain on the evolution of ebooks offers some clarity to how the shift to digital left ownership behind in the analog world.

On May 3, 2023, the White House Office of Science and Technology Policy (OSTP) requested public input on automated worker surveillance and management systems. In this Comment, the Knowing Machines Research Project (Knowing Machines) urges OSTP to translate workers’ expectations of privacy in their data into guidance for employers on when and what types of data they can collect. As worker data fuels automated surveillance technologies by serving as training data for machine-learning models, we encourage OSTP to set a high bar for employers who exploit worker data to inform employment decision and undermine workers’ autonomy. Specifically, we propose OSTP collaborate with other federal agencies to set baseline protections over worker data that align with the limits on health data for healthcare providers (HIPAA) and on consumer financial data for financial institutions respectively. At a minimum, we hope OSTP will adopt clear policies protecting worker data relating to union organizing communications and activities.Our main point: "There is growing concern about the misuse of data to train machine-learning models powering automated surveillance technologies, and worker data should be no exception.42 In its consideration of automated worker surveillance systems, OSTP must be mindful of the enclosure of worker data that enables the development of these systems in the first place. The lack of clear privacy protections for worker data provides OSTP a unique opportunity to guide employers on when and what types of worker data they can collect, store, use, and sell. OSTP must act now to counterbalance employers’ insatiable thirst for more comprehensive and invasive worker data to fuel people analytics solutions, bringing autonomy over worker data explicitly into the conversation."

This Amici Brief was filed before the Superior Court of the State of California, County of Alameda in the case of Renderos et al. v. Clearview AI, Inc. et al. in support of Plaintiffs’ opposition to Defendant Clearview’s Special Motion to Strike Pursuant to California Code of Civil Procedure § 425.16 (California’s anti-SLAPP statute).For over a century, the right of publicity (ROP) has protected individuals from unwanted commercial exploitation of their identities. Originating around the turn of the twentieth century in response to the newest image-appropriation technologies of the time, the ROP has continued to evolve to cover each new wave of technologies enabling companies to exploit peoples’ identities as part of their business models. The latest example of such a technology is Defendant Clearview AI’s facial recognition (FR) application. Clearview boasts that the primary economic value of its app stems from commercially exploiting its massive facial image database, filled with millions of individual likenesses and identities that it appropriated through images scraped from across the internet. Clearview’s misappropriations also extend to training its algorithm, matching identities to new images, and displaying results to customers. The purpose of Clearview’s product is to allow customers to identify an individual using only a picture of their face. Without the capacity to exploit millions of likenesses and identities, Clearview’s system would fail to function as a commercial product.Clearview attempts to avoid ROP liability by arguing (1) that it cannot be liable because humans rarely witness its acts of misappropriation and (2) that its app and business strategy are forms of protected speech under the First Amendment. In this brief, Amici Science, Legal, and Technology Scholars urge the Court to reject Clearview’s arguments and allow Plaintiffs’ ROP claim to proceed. First, Amici describe how the ROP claim against Clearview's FR technology is consistent with those upheld by the courts for over a century, tracing the parallel evolutions of early image-appropriation technologies and of the ROP as a legal limitation on their capacity to exploit identities for profit. Amici then apply each ROP element to Clearview’s FR app. Second, Amici challenge Clearview’s claim to protection under the anti-SLAPP statute. Clearview does not appropriate images and identities as a form of speech in connection with a public issue. Clearview is a visual surveillance company that built its app off misappropriated images for the exclusive purpose of selling and operating its commercial surveillance services, using proprietary software that it attempts to keep as far from public scrutiny as possible.If the Court finds this case is insulated from judicial review, a company can appropriate billions of individuals’ images and identities without consent, enmesh those identities in its product, license that product widely, profit lavishly, and continue with business as usual. As new products emerge that similarly undermine one’s ability to control who can use their identity and how, individuals will have less legal recourse than their ancestors had a century ago.Faced with these facts, this Court should reject Clearview's anti-SLAPP Motion and find Plaintiffs have alleged a legally valid ROP claim at this early stage. * This brief was prepared with the help of NYU Law clinical students Rupali Srivastava and Elly Brinkley under the supervision of Professor Jason Schultz, and Research Assistants Chanique Vassell, Clarie Ewing-Nelson, and Rodrigo Canalli under the supervision of the Brief authors.

It has long been recognized in Europe and elsewhere that standards-development organizations (SDOs) may adopt policies that require their participants to license patents essential to the SDO’s standards (standards-essential patents or SEPs) to manufacturers of standardized products (“implementers”) on a royalty-free (RF) basis. This requirement contrasts with SDO policies that permit SEP holders to charge implementers monetary patent royalties, sometimes on terms that are specified as “fair, reasonable and nondiscriminatory” (FRAND). As demonstrated by two decades of intensive litigation around the world, FRAND royalties have given rise to intractable disputes regarding the manner in which such royalties should be calculated and adjudicated. In contrast, standards distributed on an RF basis are comparatively free from litigation and the attendant transaction costs. Accordingly, numerous SDOs around the world have adopted RF licensing policies and many widely adopted standards, including Bluetooth, USB, IPv6, HTTP, HTML and XML, are distributed on an RF basis. This note briefly discusses the commercial considerations surrounding RF standards, the relationship between RF standards and open source software (OSS) and the SDO policy mechanisms – including “universal reciprocity” -- that enable RF licensing to succeed in the marketplace.

For over a century, the right of publicity (ROP) has protected individuals from unwanted commercial exploitation of their images and identities. Originating around the turn of the Twentieth Century in response to the newest image-appropriation technologies of the time, including portrait photography, mass-production packaging, and a ubiquitous printing press, the ROP has continued to evolve along with each new wave of technologies that enable companies to exploit peoples’ images and identities for commercial gain. Over time, the ROP has protected individuals from misappropriation in photographs, films, advertisements, action figures, baseball cards, animatronic robots, video game avatars, and even digital resurrection in film sequels. Critically, as new technologies gained capacity for mass appropriation, the ROP expanded to protect against these practices.The newest example of such a technology is facial recognition (FR). Facial recognition systems derive their primary economic value from commercially exploiting massive facial image databases filled with millions of individual likenesses and identities, often obtained without sufficient consent. Such appropriations go beyond mere acquisition, playing critical roles in training FR algorithms, matching identities to new images, and displaying results to users. Without the capacity to appropriate and commercially exploit these images and identities, most FR systems would fail to function as commercial products.In this article, I develop a novel theory for how ROP claims could apply to FR systems and detail how their history and development, both statutory and common law, demonstrate their power to impose liability on entities that conduct mass image and identity appropriation, especially through innovative visual technologies. This provides a robust framework for FR regulation while at the same time balancing issues of informed consent and various public interest concerns, such as compatibility with copyright law and First Amendment-protected news reporting.

Datasets are central to training machine learning (ML) models. The ML community has recently made significant improvements to data stewardship and documentation practices across the model development life cycle. However, the act of deprecating, or deleting, datasets has been largely overlooked, and there are currently no standardized approaches for structuring this stage of the dataset life cycle. In this paper, we study the practice of dataset deprecation in ML, identify several cases of datasets that continued to circulate despite having been deprecated, and describe the different technical, legal, ethical, and organizational issues raised by such continuations. We then propose a Dataset Deprecation Framework that includes considerations of risk, mitigation of impact, appeal mechanisms, timeline, post-deprecation protocols, and publication checks that can be adapted and implemented by the ML community. Finally, we propose creating a centralized, sustainable repository system for archiving datasets, tracking dataset modifications or deprecations, and facilitating practices of care and stewardship that can be integrated into research and publication processes.

This Amicus Brief was filed in the United States Supreme Court in the case of Google v. Oracle in support of Google's 2019 petition for certiorari.Competition and innovation are two principles at the heart of a healthy internet and the field of software development that fuels it. For decades, software engineers have relied heavily on reuse and reimplementation of functional protocols, such as the Application Programming Interfaces (APIs) in this case, to create competing alternatives to incumbent industry players and new markets for development without fear of copyright infringement. In accord with the Supreme Court’s ruling in Baker v. Selden, 101 U.S. 99 (1879), and the plain language of 17 U.S.C. § 102(b) (2012), the software industry has flourished utilizing this approach to make internet and software ecosystems more accessible, affordable, diverse, and robust.By reversing this rule in the context of APIs, the Federal Circuit upended decades of industry practice and the well-established expectations of developers, investors, and consumers. API reimplementation is a common theme among developers of all sizes — from those wishing to create entirely new platforms to those wishing to develop on them. The court below heedlessly unraveled this reasonably predictive rule and set of reliable norms that are critical to software coders for understanding what is appropriate to carry over from one project to another and what is not. This is especially true for individual coders, small startups, or nonprofit software projects, who often lack legal counsel or large financial reserves to defend themselves against unwarranted litigation.In this brief, Amici urge the Court to grant Google’s petition for certiorari in order to correct this misreading of copyright law. Specifically, Amici wish to highlight two fundamental concerns with the lower court’s opinion. First, the court’s dramatic expansion of copyright protection to include APIs, which Amici believe are not copyrightable under U.S. law, stifles innovation and competition by privileging powerful incumbents and creating artificial barriers to entry for new players and innovators where none existed before. Second, the Federal Circuit’s rejection of the fair use doctrine stands to undermine not only reimplementation and reuse of APIs, but also other valuable software engineering practices, such as reverse engineering, interoperability, and the creation of competing platforms, as well as innovations in data analytics, search engines, and many other groundbreaking advancements. Specifically, by creating irreconcilable conflicts with bedrock software fair use principles that have set the norms of engineering practice for over two decades, the Federal Circuit has opened the door to re-litigating many status quo software engineering practices — practices that open source projects and small startups depend on every day to produce new platforms, programs, features, and interfaces.This brief was prepared with the help of NYU Law clinical students Iman Charania and Michael Pizzi under the supervision of Professor Jason Schultz.

Many legal scholars have explored how courts can apply legal doctrines, such as procedural due process and equal protection, directly to government actors when those actors deploy artificial intelligence (AI) systems. But very little attention has been given to how courts should hold private vendors of these technologies accountable when the government uses their AI tools in ways that violate the law. This is a concerning gap, given that governments are turning to third-party vendors with increasing frequency to provide the algorithmic architectures for public services, including welfare benefits and criminal risk assessments. As such, when challenged, many state governments have disclaimed any knowledge or ability to understand, explain, or remedy problems created by AI systems that they have procured from third parties. The general position has been "we cannot be responsible for something we don't understand." This means that algorithmic systems are contributing to the process of government decisionmaking without any mechanisms of accountability or liability. They fall within an accountability gap. In response, we argue that courts should adopt a version of the state action doctrine to apply to vendors who supply AI systems for government decisionmaking. Analyzing the state action doctrine's public function, compulsion, and joint participation tests, we argue that-much like other private actors who perform traditional core government functions at the behest of the state-developers of AI systems that directly influence government decisions should be found to be state actors for purposes of constitutional liability. This is a necessary step, we suggest, to bridge the current AI accountability gap.

From the Pinkerton private detectives of the 1850s, to the closed-circuit cameras and email monitoring of the 1990s, to new apps that quantify the productivity of workers, and to the collection of health data as part of workplace wellness programs, American employers have increasingly sought to track the activities of their employees. Starting with Taylorism and Fordism, American workers have become accustomed to heightened levels of monitoring that have only been mitigated by the legal counterweight of organized unions and labor laws. Thus, along with economic and technological limits, the law has always been presumed as a constraint on these surveillance activities. Recently, technological advancements in several fields-big data analytics, communications capture, mobile device design, DNA testing, and biometrics-have dramatically expanded capacities for worker surveillance both on and off the job. While the cost of many forms of surveillance has dropped significantly, new technologies make the surveillance of workers even more convenient and accessible, and labor unions have become much less powerful in advocating for workers. The American worker must now contend with an all-seeing Argus Panoptes built from technology that allows for the trawling of employee data from the Internet and the employer collection of productivity data and health data, with the ostensible consent of the worker. This raises the question of whether the law still remains a meaningful avenue to delineate boundaries for worker surveillance. In this Article, we start from the normative viewpoint that the right to privacy is not an economic good that may be exchanged for the opportunity for employment. We then examine the effectiveness of the law as a check on intrusive worker surveillance, given recent technological innovations. In particular, we focus on two popular trends in worker tracking-productivity apps and worker wellness programs-to argue that current legal constraints are insufficient and may leave American workers at the mercy of 24/7 employer monitoring. We consider three possible approaches to remedying this deficiency of the law: (1) a comprehensive omnibus federal information privacy law, similar to approaches taken in the European Union, which would protect all individual privacy to various degrees regardless of whether or not one is at work or elsewhere and without regard to the sensitivity of the data at issue; (2) a narrower, sector-specific Employee Privacy Protection Act (EPPA), which would focus on prohibiting specific workplace surveillance practices that extend outside of work-related locations or activities; and (3) an even narrower sector and sensitivity-specific Employee Health Information Privacy Act (EHIPA), which would protect the most sensitive type of employee data, especially those that could arguably fall outside of the Health Insurance Portability and Accountability Act's (HIPAA) jurisdiction, such as wellness and other data related to health and one's personhood.

The entitlement to sell or dispose of one’s ownership interest in a lawful copy of a copyrighted work has long been recognized by both Congress and the courts. This “first sale” entitlement, established by a line of cases stretching back to the nineteenth century and expressly reaffirmed by Congress, embraces not only property dispositions such as resale or donation, but also limited acts of alteration and reproduction necessary to facilitate disposition. While advances in digital technology have complicated the first sale doctrine’s application, they cannot eradicate it as the District Court suggested. In fact, Congress understood the need for a copyright law that adapts to changes in technology and nothing in 17 U.S.C. § 109(a) indicates otherwise. If allowed to stand, the opinion below will undermine copyright law’s careful balance between the rights of copyright holders and consumers, leading to unanticipated and absurd results contrary to settled law and reasonable consumer expectations. For example, the District Court’s rule would require an owner of a digital song to sell her $800 smartphone to alienate a single $1 song. It would be as if selling a book at a garage sale required you to sell your bookshelf and all its contents along with it. After all, the smartphone contains not only that $1 song, but likely thousands more, along with personal photos, communications, and financial records. This outcome contradicts the plain language of § 109(a) and the longstanding history of copyright exhaustion that underlies it. Moreover, it is inconsistent with the legislative history of the Copyright Act, the property rights of digital consumers, and aspects of the District Court’s own holding. Finally, the District Court ignored the fact that the purpose of any incidental copies created by the ReDigi system was to facilitate the first sale rights consumers are entitled to effectuate under § 109(a). As a result, the District Court’s fair use analysis is fatally flawed. * NYU Technology Law & Policy Clinic Teaching Fellow Amanda Levendowski and students Cassandra Deskus, Kristen Iglesias, and Ari Lipsitz contributed heavily to this brief.

This chapter probes the reasons for consumers’ eagerness to embrace the digital marketplace despite the sacrifice of their ownership rights. It argues that the preference for digital goods is at least partially caused by consumers’ lack of information about what rights they are acquiring in digital goods. Digital retailers frequently employ misleading languages such as “buy,” “own,” “purchase” or “sell.” The chapter reports the results of a study that reveals consumers consistently overestimate their rights in digital goods, and that they prefer to purchase goods that allow them to exercise rights commonly associated with ownership. Many consumers have misconceptions about whether they can resell, lend or gift their digital goods purchased through the “Buy Now” button. A better approach should be to use short notices that unequivocally inform consumers what they are getting.

This chapter examines the battle for control between patent owners and purchasers of patented devices. By proclaiming users who violated the use restrictions as patent infringers, patent owners impose various kinds of conditions on purchasers’ use of their devices. The Supreme Court adhered to the principle of patent exhaustion and held in an 1852 case that once a patented good was sold, the patent owner could not interfere with the rights of purchasers to use it in the “ordinary pursuits of life.” While the Court stood firmly behind this principle for over 150 years, in recent years, with technological innovations like self-replicating technologies and with the development of international commerce, the common law principle of patent exhaustion is under attack. It remains to be seen how the Supreme Court will address the issue of patent exhaustion in future cases.

If you buy a book at the bookstore, you own it. You can take it home, scribble in the margins, put in on the shelf, lend it to a friend, sell it at a garage sale. But is the same thing true for the ebooks or other digital goods you buy? Retailers and copyright holders argue that you don't own those purchases, you merely license them. That means your ebook vendor can delete the book from your device without warning or explanation—as Amazon deleted Orwell's 1984 from the Kindles of surprised readers several years ago. These readers thought they owned their copies of 1984. Until, it turned out, they didn't. In The End of Ownership, Aaron Perzanowski and Jason Schultz explore how notions of ownership have shifted in the digital marketplace, and make an argument for the benefits of personal property.Of course, ebooks, cloud storage, streaming, and other digital goods offer users convenience and flexibility. But, Perzanowski and Schultz warn, consumers should be aware of the tradeoffs involving user constraints, permanence, and privacy. The rights of private property are clear, but few people manage to read their end user agreements. Perzanowski and Schultz argue that introducing aspects of private property and ownership into the digital marketplace would offer both legal and economic benefits. But, most important, it would affirm our sense of self-direction and autonomy. If we own our purchases, we are free to make whatever lawful use of them we please. Technology need not constrain our freedom; it can also empower us.

This chapter begins by considering the emergence of the sharing economy, which brings consumers cheaper access and convenience. But it also deprives consumers of some rights that are commonly associated with property ownership. This chapter then proposes several ways to combat the erosion of ownership. Legal reform faces many hurdles, but there are still several steps legislators and courts can take: first, false claims of ownership such as “buy now” or “own” should be targeted to improve consumer information; second, the pervasive use of form contracts needs to be curbed; third, consumers need to be freed from the restrictions imposed by DRM; lastly, courts also need to reinvigorate the principle of patent exhaustion and reform copyright law. Technological developments, including using block chain technology to record ownership interests in digital assets, can also help ensure meaningful personal property rights.

This chapter outlines the conceptual framework for the rest of the book by describing the basic principles of personal and intellectual property law. The default rules of ownership used to define the purchasers’ rights, allowing for resale, lending, gifting and many other forms of transaction by the consumer. Now, with the rise of the digital economy, consumer rights are defined through licenses, which impose various restrictions on consumers’ disposition of their digital goods. Among the traditional rules of ownership, the exhaustion principle is particularly important for mediating the tension between intellectual property holders and consumers. Exhaustion is the notion that an IP rights holder relinquishes some control over a product once it sells or gives that product to a new owner. IP rights holders have resisted exhaustion at nearly every turn, and the licensing model allowed them to infuse new vigor into their resistance.

This chapter examines the license agreements imposed by IP rights holders that redefine transactions and strip consumers of ownership even after an apparent sale. Despite their importance, consumers seldom read these license agreements because of their length and complexity. In response, IP rights holders produce highly uniform license terms that impose restrictions on the rights acquired by consumers. There are two approaches of interpreting license agreements: one treating them as contracts that require the mutual consent to be effective, while the other construing license agreements as expression of permission that does not require agreement to be effective. Many courts rely on license agreements to determine whether consumers enjoy ownership over the things they purchase. The better approach, however, should be to look at the economic reality of a transaction.