Discussions regarding the dual use of foundation models and the risks they pose have overwhelmingly focused on a narrow set of use cases and national security directives-in particular, how AI may enable the efficient construction of a class of systems referred to as CBRN: chemical, biological, radiological and nuclear weapons. The overwhelming focus on these hypothetical and narrow themes has occluded a much-needed conversation regarding present uses of AI for military systems, specifically ISTAR: intelligence, surveillance, target acquisition, and reconnaissance. These are the uses most grounded in actual deployments of AI that pose life-or-death stakes for civilians, where misuses and failures pose geopolitical consequences and military escalations. This is particularly underscored by novel proliferation risks specific to the widespread availability of commercial models and the lack of effective approaches that reliably prevent them from contributing to ISTAR capabilities. In this paper, we outline the significant national security concerns emanating from current and envisioned uses of commercial foundation models outside of CBRN contexts, and critique the narrowing of the policy debate that has resulted from a CBRN focus (e.g. compute thresholds, model weight release). We demonstrate that the inability to prevent personally identifiable information from contributing to ISTAR capabilities within commercial foundation models may lead to the use and proliferation of military AI technologies by adversaries. We also show how the usage of foundation models within military settings inherently expands the attack vectors of military systems and the defense infrastructures they interface with. We conclude that in order to secure military systems and limit the proliferation of AI armaments, it may be necessary to insulate military AI systems and personal data from commercial foundation models.

Ongoing political, environmental, and economic crises require infrastructures that can respond to crises in ways that do not replicate and reinforce inequality. To this end, we use a case study method of analysis that compares the authors' previous work on Internet infrastructure at the levels of development, governance, and use to explore how these imaginaries promote or impede people-centered change in the development and maintenance of Internet infrastructure. This theoretical work puts the three existing cases in conversation to better understand how Internet infrastructure alternatives presented as radical, new, or non-hierarchical present shortcomings and opportunities, so that it might be more possible to imagine better, more truly radical, people-centered alternatives. From this comparison, we close our discussion with three heuristics for radical infrastructure: the need for pushing for alternative ensembles of support, busting the myth of technosolutionism, re-politicizing Internet infrastructure, and encouraging technical communities to build around cooperativity, not connectivity.

This paper examines 'open' artificial intelligence (AI). Claims about 'open' AI often lack precision, frequently eliding scrutiny of substantial industry concentration in large-scale AI development and deployment, and often incorrectly applying understandings of 'open' imported from free and open-source software to AI systems. At present, powerful actors are seeking to shape policy using claims that 'open' AI is either beneficial to innovation and democracy, on the one hand, or detrimental to safety, on the other. When policy is being shaped, definitions matter. To add clarity to this debate, we examine the basis for claims of openness in AI, and offer a material analysis of what AI is and what 'openness' in AI can and cannot provide: examining models, data, labour, frameworks, and computational power. We highlight three main affordances of 'open' AI, namely transparency, reusability, and extensibility, and we observe that maximally 'open' AI allows some forms of oversight and experimentation on top of existing models. However, we find that openness alone does not perturb the concentration of power in AI. Just as many traditional open-source software projects were co-opted in various ways by large technology companies, we show how rhetoric around 'open' AI is frequently wielded in ways that exacerbate rather than reduce concentration of power in the AI sector.

The contemporary Internet's "network of networks" has become infrastructural to our lives. The Internet is a stack of physical, data link, network, transport, and application layers which all have unique rules and roles. While many see Internet infrastructure as a foregone conclusion, Paris, Cath and Myers West (2023) write “Internet infrastructure is built slowly, over time, protocol by protocol, in response to many different technical, social, political, environmental, and economic imperatives”. Even as the particular model of the Internet we are all accustomed to has become the standard, other attempts proliferated and eventually failed, as did the Soviet Internet (Peters 2016), and as this panel highlights, the Internet is still ever-evolving. The project of this panel is to trace alternative, parallel, and emergent network models, standards and protocols, theorize their impact as they appear in different places, spaces, and contexts, and gesture towards how the Internet might be different. As critical internet studies have since the early 2000s shown, computational standards, protocols, and network diagrams are more than technical details, they have the power to shape and structure the conditions for our socio-cultural lifeworlds (Galloway 2006; Chun 2008; Bratton 2016). As Gehl (2014) puts it: “interfaces, database structures, mechanisms of connection all shape social activities”. Change an element in the stack and a different connectivity, a different future becomes possible. The papers of this panel introduce and discuss five different and potentially revolutionary network technologies that manage and organize our online lives. The first paper represents a media genealogy of ActivityPub – a protocol that enables the Fediverse, a collection of social media sites that can communicate with one another. The author argues that ActivityPub was not produced through an instrumental process, but was the result of accidents and coincidences. The accidental nature of the protocol, coupled with its being authored by self-identified queer and trans developers, has put it on a collision course with both the “standard” approach to standards production as well as mainstream, corporate social media. The second paper focuses on the design of the Interplanetary Internet and the idea of delay-tolerant networking fundamental to operating in outer space. The author maintains that when delays are central to a network model, we are forced to rethink how our connections are maintained and organized in the future. Delay-tolerant networking is thus not only a technical solution for a communications system but a control protocol through which interplanetary life can be managed. The third paper is also focused on the temporality of networks. The third paper examines how time is enacted as a design ideology in the course of the development of a future internet architecture protocol project: named data networking (NDN). This work locates aspects of the sociomateriality of time in the processes of building Internet infrastructure and demonstrates how it binds together cultural, economic, and discursive power. The paper argues that thinking through time as a design ideology can be useful in projects imagining how the Internet might be built to engender and support different values than market ideology. The fourth paper is about the organizational culture of the Internet Engineering Task Force (IETF), a key internet standards and protocol organization. The paper argues that the organization is guided by a culturally inflected anti-political engineering ethos, whose depoliticizing tendencies hampers the organization’s functioning and its ability to rise above narrow industry-interest and pursue a public interest internet. The fifth paper looks to the Crypto Wars of the 1990s as a moment where things could have been otherwise; comparing the examples of PGP and RSA encryption software and how they shaped the nature of our networked systems. It argues that a combination of regulatory and commercial interests influenced the development and use of cryptography in ways that facilitated the development of e-commerce, but left private messaging in dubious legal status. Collectively the papers investigate alternative and emergent trends behind the Internet and its network models, standards, and protocols. The protocols and rules for network connection, standards bodies, and modes of governance are critical to maintaining and upkeeping a network. Their impact, however, is not merely technical but potentially world-changing. The papers direct their critical gaze towards the development of these technologies and what their introduction to our world potentially entails. By focusing on projects of past, present, and future and by exploring the Internet’s deepest sociotechnical layers, the panel critically dismantles the commonly-held idea that the Internet is a monolith and illustrates that the history of the Internet is still being written.

Digital advertising and technology companies are resigned to a new privacy imperative. They are bracing for a world where third-party tracking will be restricted by design or by law. Digital resignation typically refers to how companies cultivate a sense of powerlessness about privacy among internet users. Our paper looks through this optic from the other end of the lens: How is the digital advertising industry coping with the increasing salience of privacy? Recent developments have forced companies to implement “privacy-preserving” designs—or at least promise some semblance of privacy. Yet, the industry remains dependent on flows of data and means of identification to enable still-desired targeting, measurement, and optimization. Our paper analyzes this contradiction by looking at systems that aim to replicate existing functionalities while protecting user “privacy.” We call this a form of “cynical resignation” and characterize its key maneuvers as follows: (a) sanitizing surveillance ; (b) party-hopping ; and (c) sabotage . We argue that this “cynical resignation” to a privacy imperative represents a policy failure. In the absence of decisive interventions into the underlying business models of data capitalism, companies offer techno-solutionism and self-regulations that seem to conform to new laws and norms while reinforcing commitments to data-driven personalization. This may benefit the largest tech companies, since their privileged access to first-party data will make more companies reliant on them, and their computational power will be even more valuable in a world where modeling is used to compensate for the loss of third-party data and traditional methods of personal identification.

This paper examines ‘open’ AI in the context of recent attention to open and open source AI systems. We find that the terms ‘open’ and ‘open source’ are used in confusing and diverse ways, often constituting more aspiration or marketing than technical descriptor, and frequently blending concepts from both open source software and open science. This complicates an already complex landscape, in which there is currently no agreed on definition of ‘open’ in the context of AI, and as such the term is being applied to widely divergent offerings with little reference to a stable descriptor. So, what exactly is ‘open’ about ‘open’ AI, and what does ‘open’ AI enable? To better answer these questions we begin this paper by looking at the various resources required to create and deploy AI systems, alongside the components that comprise these systems. We do this with an eye to which of these can, or cannot, be made open to scrutiny, reuse, and extension. What does ‘open’ mean in practice, and what are its limits in the context of AI? We find that while a handful of maximally open AI systems exist, which offer intentional and extensive transparency, reusability, and extensibility– the resources needed to build AI from scratch, and to deploy large AI systems at scale, remain ‘closed’—available only to those with significant (almost always corporate) resources. From here, we zoom out and examine the history of open source, its cleave from free software in the mid 1990s, and the contested processes by which open source has been incorporated into, and instrumented by, large tech corporations. As a current day example of the overbroad and ill-defined use of the term by tech companies, we look at ‘open’ in the context of OpenAI the company. We trace its moves from a humanity-focused nonprofit to a for-profit partnered with Microsoft, and its shifting position on ‘open’ AI. Finally, we examine the current discourse around ‘open’ AI–looking at how the term and the (mis)understandings about what ‘open’ enables are being deployed to shape the public’s and policymakers’ understanding about AI, its capabilities, and the power of the AI industry. In particular, we examine the arguments being made for and against ‘open’ and open source AI, who’s making them, and how they are being deployed in the debate over AI regulation. Taken together, we find that ‘open’ AI can, in its more maximal instantiations, provide transparency, reusability, and extensibility that can enable third parties to deploy and build on top of powerful off-the-shelf AI models. These maximalist forms of ‘open’ AI can also allow some forms of auditing and oversight. But even the most open of ‘open’ AI systems do not, on their own, ensure democratic access to or meaningful competition in AI, nor does openness alone solve the problem of oversight and scrutiny. While we recognize that there is a vibrant community of earnest contributors building and contributing to ‘open’ AI efforts in the name of expanding access and insight, we also find that marketing around openness and investment in (somewhat) open AI systems is being leveraged by powerful companies to bolster their positions in the face of growing interest in AI regulation. And that some companies have moved to embrace ‘open’ AI as a mechanism to entrench dominance, using the rhetoric of ‘open’ AI to expand market power while investing in ‘open’ AI efforts in ways that allow them to set standards of development while benefiting from the free labor of open source contributors.

Whether or not they've embraced this role, social media companies have long played an important role in shaping international politics. This chapter examines the political economy of social media companies by situating them as geopolitical actors. Drawing on several case studies, and emphasizing examples from outside of the US and EU, it provides an overview of several distinct issues posed by social media in different parts of the world. It concludes with a critical examination of the notion of social media companies as 'social infrastructure' - and how we can build beyond critique toward collective action to improve their accountability.

In 1976, two researchers declared a revolution in cryptography: With the invention of public key encryption, cryptography could be used not only to share secret messages, but to secure and authenticate communications networks, and, eventually, to enable radically new kinds of social relationships facilitated by networked communication technology. This article explores a series of transformations in the meaning of cryptography in the 1960s and 1970s that led to the declaration of a revolution. Drawing on archival materials, the article considers how public key cryptography was the product of an emerging consensus among cryptographers of the importance of privacy in the wake of abuses of surveillance powers by government agencies. Shaped by a changing technological and political environment, it situates cryptography at the center of a focused effort to assert control over information in an era of sociopolitical upheaval, concluding that the invention of public key encryption both marked a change in the imaginary surrounding cryptography and offered a technical solution that foreclosed other approaches to addressing the problem of surveillance.

This article explores an inflection point for a community of cryptography advocates as they grappled with a series of cascading failures. Drawing on 3 years of ethnographic observation and interviews at conferences devoted to building privacy systems, I consider how a determinist conception of encryption technologies inhibited the widespread adoption of privacy technologies. I develop the frame of “survival of the cryptic” to call attention to the way this conception fails to acknowledge how power shapes the conditions of surveillance: that race and racism, gender and misogyny affect not only who is most impacted by surveillance but also how the encryption technologies developed to inhibit surveillance were designed—and, as importantly, who they were designed for. I conclude by offering a new imaginary for encryption that draws on queer, black and feminist thought by centering the need to create safe and autonomous spaces for collective survival under conditions of mass surveillance.

Content moderation has exploded as a policy, advocacy, and public concern. But these debates still tend to be driven by high-profile incidents and to focus on the largest, US based platforms. In order to contribute to informed policymaking, scholarship in this area needs to recognise that moderation is an expansive socio-technical phenomenon, which functions in many contexts and takes many forms. Expanding the discussion also changes how we assess the array of proposed policy solutions meant to improve content moderation. Here, nine content moderation scholars working in critical internet studies propose how to expand research on content moderation, with implications for policy.

Computer scientists, and artificial intelligence researchers in particular, have a predisposition for adopting precise, fixed definitions to serve as classifiers (Agre, 1997; Broussard, 2018). But classification is an enactment of power; it orders human interaction in ways that produce advantage or suffering (Bowker & Star, 1999). In so doing, it obscures the messiness of human life, masking the work of the people involved in training machine learning systems, and hiding the uneven distribution of its impacts on communities (Taylor, 2018; Gray, 2019; Roberts, 2019). Feminist scholars, and particularly feminist scholars of color, have made powerful critiques of the ways in which artificial intelligence systems formalize, classify, and amplify historical forms of discrimination and act to reify and amplify existing forms of social inequality (Eubanks, 2017; Benjamin, 2019; Noble, 2018). In response, the machine learning community has begun to address claims of algorithmic bias under the rubric of fairness, accountability, and transparency. But in doing so, it has largely dealt with these issues in familiar terms, using statistical methods aimed at achieving parity and deploying fairness ‘toolkits’. Yet actually existing inequality is reflected and amplified in algorithmic systems in ways that exceed the capacity of statistical methods alone. This article outlines a feminist critique of extant methods of dealing with algorithmic discrimination. I outline the ways in which gender discrimination and erasure are built into the field of AI at a foundational level; the product of a community that largely represents a small, privileged, and male segment of the global population (Author, 2019). In so doing, I illustrate how a situated mode of inquiry enables us to more closely examine a feedback loop between discriminatory workplaces and discriminatory systems.

On March 28, 2019, the AI Now Institute at New York University (NYU), the NYU Center for Disability Studies, and Microsoft convened disability scholars, AI developers, and computer science and human-computer interaction researchers to discuss the intersection of disability, bias, and AI, and to identify areas where more research and intervention are needed. 1This report captures and expands on some of the themes that emerged during discussion and debate. In particular, it identifies key questions that a focus on disability raises for the project of understanding the social implications of AI, and for ensuring that AI technologies don’t reproduce and extend histories of marginalization.

This article seeks to provide greater specificity to demands for transparency in the commercial content moderation practices of digital platforms. We identify gaps in knowledge through a thematic analysis of 380 survey responses from individuals who have been the subject of content moderation decisions. We argue that meaningful transparency should be understood as a component of a communicative process of accountability (rendering account) to independent stakeholders. We make specific recommendations for platforms to provide people with clear information about decisions that affect them, including what content is moderated, which rule was breached, and a description of the people and automated processes responsible for identifying content and making the decision. Beyond providing more information to individuals about particular decisions, however, we note the major challenge of improving understanding of content moderation at a systems level. General demands for greater transparency should be reframed to focus on enhanced access to large-scale disaggregated data that can enable new methods and collaborations among academia, civil society, and journalists to make these systems more understandable and accountable.

The role of the public in Internet governance debates is a critical issue for policymaking, even more so as Internet governance forums encounter crises of legitimacy. This article examines the role of the public in Internet governance debates, drawing connections between theories of the public, science and technology studies, and Internet governance. It examines the architecture of public inclusion at the Global Multistakeholder Meeting on the Future of Internet Governance (NETmundial). NETmundial emerged at a point when public legitimacy became particularly salient for the Internet governance community. So finding modes of inclusion for the rapidly growing, and increasingly diverse group of stakeholders in the governance debate was especially important for the meeting's success. This article links the debate to notions of the public sphere—both in shaping a discursive space central to public debate, the Internet, and by championing a particular form of public discourse, multistakeholderism. By drawing out these connections, the article both reflects upon and challenges ideas about how to achieve public inclusion in Internet governance.

This article provides an overview of the key values that we argue should underpin an index of the legitimacy of the governance of online intermediaries. The aim is ultimately to allow scholars to rank the policies and practices of intermediaries against core human rights values and principles of legitimate governance in a way that enables comparison across different intermediaries and over time. This work builds on the efforts of a broad range of researchers already working to systematically investigate the governance of social media platforms and telecommunications intermediaries. In this article, we present our review and analysis of the work that has been carried out to date, using the digital constitutionalism literature to identify opportunities for further research and collaboration.

This paper interrogates discourses associated with encryption in contemporary policy debates. It traces through three distinct cryptographic imaginaries – the occult, the state, and democratic values – and how each conceptualises what encryption is, what it does, and what it should do. Situating each imaginary in time through historical research, I consider how they foreground distinct configurations of power and authority. It concludes by describing the development of a new cryptographic imaginary, one which sees encryption as a necessary precondition for the formation of networked publics.

Social media platforms play an increasingly important civic role as platforms for discourse, where we discuss, debate, and share information. This article explores how users make sense of the content moderation systems social media platforms use to curate this discourse. Through a survey of users ( n = 519) who have experienced content moderation, I explore users’ folk theories of how content moderation systems work, how they shape the affective relationship between users and platforms, and the steps users take to assert their agency by seeking redress. I find significant impacts of content moderation that go far beyond the questions of freedom of expression that have thus far dominated the debate. Raising questions about what content moderation systems are designed to accomplish, I conclude by conceptualizing an educational, rather than punitive, model for content moderation systems.

Scholars face serious difficulties in gaining access to examine the practices of content moderation within commercial platforms. Over the last few years, we have seen a number of excellent detailed qualitative analyses of governance practices of particular platforms, but much remains shrouded in secrecy. It is difficult, in this context, for scholars researching the broader social implications of commercial content moderation practices to access the information they need to study changes over time and across platforms. In this panel, we consider the opportunities for methodological innovation and cross-disciplinary collaboration to help progress future research. Speakers include: Sarah Myers West: User-focused studies, survey/interview based research Nicolas Suzor: Webscraping, APIs, and transparency reports: developing a better picture of the practices of content moderation at scale Nathalie Marechal: Ranking Digital Rights and measuring the human rights impact of content moderation policies Sarah Roberts - studying content moderation and digital labor