The rapid expansion of artificial intelligence (AI) technologies has intensified debates over data sourcing, legality, and knowledge legitimacy. This paper examines the epistemic ruptures that arise when proprietary AI models are trained on data extracted from shadow libraries, focusing on Meta’s use of Anna’s Archive. As a repository of copyrighted materials operating in a legal gray zone, Anna’s Archive challenges conventional distinctions between legitimate and illegitimate knowledge. Meta’s reliance on its data underscores broader tensions in digital governance, knowledge commodification, and corporate data strategies.

Using a Critical Data Studies (CDS) framework, this study investigates how power asymmetries shape AI knowledge production and legal discourse. By analyzing the Kadrey v. Meta Platforms, Inc. case, the paper explores how judicial reasoning reinforces corporate ownership of AI-generated knowledge while marginalizing alternative epistemologies. Key findings include the legal codification of epistemic ruptures, the reinforcement of shadow library invisibility, and the ambiguous legitimacy of knowledge sourcing.

Meta’s legal maneuvers, such as sealing motions and evidence gating, highlight corporate dominance in AI data acquisition, making it difficult for challengers to contest its practices. Meanwhile, shadow libraries navigate an ideological battlefield, portraying their work as a moral imperative while facing legal and political scrutiny. This paper contributes to ongoing discussions on AI ethics, copyright law, and digital knowledge infrastructures, aligning with AoIR 2025’s theme of "ruptures" to critically examine AI’s role in reshaping access, control, and legitimacy.

Data donation makes it possible to invite participants to request and share their data from digital platforms for research purposes. While seen as a user-centric approach to digital trace data collection, little is known about how participants experience and understand the data donation process. This study therefore asked twenty participants to verbalize their thoughts and actions as they went through data donation (from Google, YouTube, Facebook, X, Instagram, TikTok, LinkedIn), and interviewed them about their experiences. Using a workflow that visualized participants’ own data and enabled them to inspect and delete data prior to donation, we not only identified strengths and obstacles within the data donation process from a user perspective, but also gained insight into how participants make sense of data donation. Overall, we find that while participants enjoyed gaining insight into their own media behavior, their understanding of data donation was problematic. Most participants misunderstood or overlooked the option to search through and delete their data, raising questions about meaningful informed consent. Another key finding was that participants interpreted data visualizations as objective representations of their media use, even when these data were incomplete or contradicted their own ideas about their platform use. Surprisingly, privacy considerations were only at play during platform selection, but not during the actual donation process.

This paper traces the consequences of the decimation of the US Agency for International Development (USAID) for a refugee camp along the border between Thailand and Myanmar. Events in Washington DC in January 2025, had devastating effects in Mae La camp as medical care and other essential services were withdrawn overnight. A less discussed dimension of the crisis concerns the potential harms resulting from the capture of refugee sensitive data. The paper analyses these events as exemplars of ‘infrastructuring’ and ‘infrastructural violence’. Digital technologies and AI increasingly underpin humanitarian operations to the extent that they provide the infrastructure through which essential aid is delivered. A typical example is biometrics which is commonly used for the delivery of vital services in Mae La. As infrastructures transcend institutional boundaries and humanitarian systems become interoperable with those of private companies and nation-states, the opportunities for function creep increase. The shareability of data and the permanence of records mean that data collected for one reason may be used for entirely different purposes and by different actors than the humanitarian organisations. The infrastructuring of aid amplifies the risks to individuals and multiplies the opportunities for structural violence.

This paper offers a critical technography of ‘audio-first’ digital platform Spotify, scrutinizing the ways in which the company’s institutional discourse and promotional materials frame the practices of datafication and personalization. Engaging with the growing literature on platform studies, critical algorithm studies, and everyday data cultures, I argue that the case of Spotify represents a paradigmatic example of a broader discursive shift in which data practices are justified by platforms based on their capacity to evoke ‘delight’. On Spotify, I identify that this phenomenon is manifested through the mobilization of a contradictory ‘computational mojo’ that combines the science of numbers with miraculous, magical powers, as well as a focus on the timeliness and beauty of its algorithmic-driven contextual harmony. In so doing, Spotify’s designed experiences attempt to convince users that they should revel in acts of ‘data indulgence’ – the craving, appetite for, and enjoyment of data-driven experiences. Rather than accepting that these data-craving and advertising-friendly modalities of subjectivity are naturally or automatically occurring, I argue that the phenomenon illustrates the complex dynamics of ‘staging power’. Dialoguing with recent debates on the role of affect, emotion, and desire in mediating datafication, I aim to contribute to, and expand on, existing theorizations on the symbolic and aesthetic power of digital platforms.