Open Science has long been framed as a normative project grounded in values of transparency, accessibility and collaboration. Yet in the age of generative AI, many research data practitioners, who once championed openness, are increasingly questioning how these commitments can be upheld when openness also enables undesired large-scale data extraction, decontextualization and enclosure. This paper examines how the normativities of openness are being re-negotiated, and how researchers navigate the ethical and political tensions that arise when their work is repurposed for proprietary AI systems.

Drawing on in-depth qualitative interviews, I explore how situated ethical reasoning informs decisions about data sharing. Participants described new risks and dilemmas, from uncontrolled scraping and heightened re-identification threats to the reframing of FAIR (Findable, Accessible, Interoperable, Reusable) principles as pipelines for AI-readiness. Their accounts reveal openness as a relational and contested practice, embedded in specific institutional, infrastructural and community contexts, and continuously re-balanced between care, accountability and collaboration.

Rather than rejecting openness outright, interviewees described recalibrating it – through selective sharing, alternative infrastructures and advocacy for commons-based governance models.

Framed by critical data studies, and science and technology studies, the analysis conceptualizes these practices as reflexive agency: the situated capacity to align openness with evolving ethical and political conditions. The findings argue for re-politicizing openness as a matter of commons stewardship and situated ethics, aligning governance, infrastructure and normative commitments to resist extractive appropriation while sustaining the collective conditions for shared knowledge production.

This article examines how artificial intelligence (AI) systems engage with archival collections as data, proposing a critical framework to interrogate archival records positioned as data for AI. I present a three-part heuristic for analysing archival records in AI contexts: first, by understanding records as processual entities; second, by examining the infrastructural power that shapes access and preservation of records in AI; and third, by focusing on obfuscation of provenance information and selective inclusion of data that creates new forms of archival silence and exclusion. By reading such data against the training window, critical AI scholars, archivists and archival researchers can better understand the limitations of engaging AI in cultural heritage, reassert institutional control over collections and develop more responsible approaches to documenting archival materials incorporated by AI systems as data resources.

Archives are essential vehicles for both historical research and memory practices like genealogy. The sheer volume of documents produced via digital infrastructures, however, has posed a challenge for traditional archival processes that rely on human judgments. Some archivists have responded to this challenge by engaging with research in natural language processing (NLP), a subfield of AI that offers practical tools for working with large corpora of text. In this paper, we review applications of NLP methods in archival processes in order to surface fundamental tensions between archival values and automation. NLP can be used to classify materials as sensitive, recognize names in documents and facilitate new kinds of search, among other applications. Our discussion of these issues follows the journey of a document through an archive, from acquisition to interpretation. We find that sensitivity to context, transparency, resource limitations and maintenance appear as recurring challenges that are complicated by automation. The benefits that NLP technologies lend to managing large archives, however, suggest that these tools should be considered, not discarded. Turning to a more forward-looking discussion of generative AI and archival infrastructure, we argue that the narrative and dialogic potential of contemporary large language models has an especially contentious relationship with conventional archival desiderata.