Artificial Intelligence presents both unprecedented opportunities and unexpected challenges for open repositories. This panel brings together voices from across the InvenioRDM community - service providers, institutional implementers, and large-scale research infrastructure - to explore how AI is reshaping repository sustainability.

AI offers genuine enhancements: intelligent submission workflows, automated curation assistance, and improved discoverability. Yet aggressive AI scraping behavior threatens the infrastructure enabling openness. At the recent CNI meeting, bot management emerged as a critical institutional concern. Repositories face scraper waves consuming resources at denial-of-service scales, preventing legitimate user access. Caltech has experienced similar challenges across InvenioRDM and EPrints platforms.

Beyond infrastructure, there's a human cost: curators battling AI-generated spam submissions, predatory content, and relentless bot traffic. CERN's recent analysis reveals the toll on repository staff.

This panel presents concrete institutional experiences, examines mitigation strategies and trade-offs, and facilitates discussion about collective responses. How do we embrace AI's benefits while managing its challenges? When does "Open to All" become unsustainable?

Attendees will gain practical insights, real-world data, and join a crucial conversation about open repositories' future in an AI-saturated landscape.

Over the past three years, the Canadian Repositories Community of Practice Steering Committee, supported by the Canadian Association of Research Libraries, has been privileged to support a growing community of practice among Canadian repository professionals. While communities of practice are helpful across many disciplines, they can be especially impactful for advancing repository infrastructure. Locally, repository managers are often isolated at their institutions, working with limited resources to try to satisfy institutional stakeholders and encourage people to share their work. However, the greatest potential for institutional repositories lies not in these individual efforts, but as infrastructure to connect scholarly communication worldwide. Working together via a community of practice is great for support, resources, and mentorship, but it also enables us to collaboratively envision the future of universal indexing (Chavarro et al., 2025) and the role we might play as interconnected nodes in a “distributed ecosystem” (COAR, 2024). Our committee work has included hosting regular events, developing a mentorship program, and participating in ongoing conversations around important topics for repositories. We are sharing our experience to inspire others to continue these conversations in their own contexts so we can work together towards interoperability through initiatives that are scalable, non-resource-intensive, and accessible.

Scholaris is a new national opt-in shared repository service that aims to support the open discovery and preservation of Canadian scholarship by providing scalable infrastructure and support for institutional repository teams.

Alongside the technical aspects, community engagement is a necessary piece of building a forward-thinking service that reflects the shared needs and interests of the Canadian repository community. Recognizing the importance of building the technical and community infrastructure in tandem, three Network of Expert Groups were convened in Spring 2024 to provide recommendations on key areas of repository management: metadata and discovery, electronic theses and dissertations, and digital preservation. Over a two-year period, each group made significant contributions to the development of Scholaris, bringing critical perspectives, diverse expertise, and enthusiasm to the initiative. While their work was grounded in the Scholaris context, the resulting outputs are broadly relevant to the global repository community.

This presentation shares insights and lessons learned from convening the expert groups and perspectives from implementing their recommendations within the service model. The session concludes with practical suggestions for navigating the complexities of developing a shared repository service and demonstrates the value of prioritizing community engagement from the outset.

When researchers submit content to a repository, they often lack the knowledge and experience required to make their submissions as reusable (and FAIR) as desired. Repository staff can greatly improve the quality by guiding researchers through the submission process, but this support can be time-consuming and costly. If the repository submission system is designed to actively assist researchers throughout the submission process, the resulting submission will have a higher level of quality before it reaches human repository staff, reducing the amount of support required to make the submission ready for acceptance.

Dryad (https://datadryad.org) has redesigned our submission process from the ground up to support researchers in submitting publishable files and describing the files with high-quality metadata. This presentation will describe the improvements we have made from our previous submission process, showcase the features that have the greatest impact on quality, and describe how we are building flexibility for future enhancements.

The presentation will be focused on the Islandora system and will be divided into three parts. In the first part we will shortly introduce Islandora itself and what features are available out of the box. The next part will show how Islandora can be extended and customized using configuration, views and modules in Islandora’s frontend (Drupal) to satisfy very specific user needs. The last part will present some of our Islandora repository instances to highlight what customizations are possible in practice, such as displaying search results on a map or comparing images and metadata.

Persistent identifiers (PIDs) like DOIs, ORCID, and ROR have improved linking between scholarly articles, authors, and institutions, enabling better interoperability. However, these PIDs alone do not fully ensure transparency in open access publishing. Many articles are published under transformative agreements, memberships, or sponsorships, but metadata about these agreements is often missing, making it hard to verify why a Version of Record is open access. To address this, a central, openly accessible registry of publisher agreements is proposed, including metadata on institutions, timelines, and agreement types. Embedding this registry in existing infrastructures like the Electronic Journal Library (EZB) would allow repositories and service providers to reuse data efficiently. The openCost project complements this by defining a metadata schema for cost data and agreements, ensuring machine-readable, open formats and compatibility with global standards like DataCite. Agreement PIDs, similar to DOIs or RORs, are essential for linking articles to agreements. Implementations in institutional repositories and services like OpenAPC demonstrate feasibility.

As arXiv celebrates its 35th anniversary in 2026, we are both proud of our accomplishments and excited about the future. For several years, arXiv has been working toward a complete technical migration which we expect to complete in June 2026. During this time, we have built out our governance structure and volunteer operations, while also experiencing record growth in submissions. We'll share our experiences and plans for addressing a rapidly changing landscape.

Digital preservation of academic journals is essential to ensure long-term access to their contents, which requires the use of durable formats that are resilient to technological change. In traditional editorial workflows, journals typically produce generic PDF files that are deposited in repositories without guarantees of long-term preservation. As a result, repositories must independently decide whether to convert these files into PDF/A or other preservation-friendly formats. Moreover, the absence of a single, widely adopted standard limits interoperability between publishers and repositories. This work proposes leveraging the publication workflow of Open Journal Systems (OJS) to generate, at the point of origin, both JATS XML and preservation-ready PDF/A files, and to package them using the BagIt standard to facilitate repository deposit. In the proposed implementation, OJS plugins automatically convert articles from DOCX to JATS and generate final PDF/A outputs. This integrated approach shifts preservation responsibilities to the editorial workflow, significantly reducing the burden on repositories while improving interoperability and archival readiness. To date, the system has successfully generated JATS XML and PDF/A files, with BagIt packaging planned as the next step toward a complete preservation workflow.

The Swiss DLCM's open access technology enables long-term digital preservation. The strengths of this technology lie in its modularity, scalability and open architecture, which comply with the OAIS standard to guarantee the integrity of the archives and mitigate preservation risks. By using the DataCite schema for metadata, assigning a persistent identifier to each archive, and using persistent identifiers in metadata for individuals, institutions and licenses, DLCM adheres to FAIR principles. An ergonomic Web portal facilitates data deposit and access based on specific permission levels and data sensitivity in the fields of research, administrative, and heritage. The ingestion phase automatically qualifies the formats submitted, informing users of any risk of file obsolescence by assigning a compliance level based on file format identification, performed by the PRONOM registry. Ready for Core Trust Seal certification, designed using cutting-edge software practices, DLCM guarantees maintainability, including compatibility with “DevOps” practices. Beyond interfacing with standard storage industry protocols, DLCM innovates with a DNA connector. This connector, developed as part of the European DNAMIC project (https://dnamic.org), includes all the processing steps necessary to encode binary files into chemically synthesized nucleotides and vice versa, and includes error correction mechanisms to recover the archives embedded in these tiny molecules.

How do you preserve and provide access to millions of previously internal corporate documents released through public interest lawsuits? Staff from the Industry Documents Library (IDL) at the University of California San Francisco will share an overview of the IDL repository, and how a box of documents from a tobacco company whistleblower launched a digital archive that now includes over 26 million freely available documents obtained from industries which impact public health. The team will discuss the challenges of navigating the legal landscape of document disclosure, managing large volumes of digitized and born-digital records which may contain sensitive information, and facilitating open access for a wide audience of academic researchers, students, policymakers, public health experts, educators, legal professionals, journalists, and community advocates. The presentation will offer tips and lessons learned for accessioning very large digital collections, sustaining staffing and funding, establishing policies and workflows for safeguarding sensitive data, and centering user needs in repository design and decision-making.

Despite early discussions of Open Educational Resources (OERs) in Sweden, their uptake remains limited, partly due to fragmented infrastructure. Recent national open science guidelines and the updated roadmap of the Association of Swedish Higher Education Institutions (SUHF) call for stronger coordination, a sharing culture, and international collaboration around OER. This abstract presents the development of a forthcoming lightweight national platform for OERs in Sweden that directly addresses these ambitions.

The initiative combines technical reuse with service design. Technically, Karlstad University and the Technische Informationsbibliothek (TIB), together with Sunet and Swedish HEIs, integrate the Open Educational Resources Search Index (OERSI) as a central search layer that aggregates distributed OER repositories without hosting content itself. A simple WordPress-based front end, open APIs, and repository-grade services such as Zenodo minimize proprietary development while ensuring persistence, openness, and FAIR-aligned discovery.

Equally important is the process behind the solution. Using design-led and agile methods, the project evolved through iterative prototyping, shared governance, and close engagement with educators, librarians, and national actors. The work was further shaped by cultivated serendipity arising from international collaboration and a “ready to reuse” mindset. The case offers a transferable reference architecture and practical lessons for sustainable, collaborative national (OER) infrastructures.

Resilience in times of crisis: Strengthening Open Science against geopolitical pressures